# Actual Database — full text
> Numbers we measured, not numbers a model guessed.

Generated: 2026-10-26

---

# Backblaze B2 vs Bunny Storage: 1 TB for a Month, Priced

URL: https://actualdatabase.com/backblaze-b2-vs-bunny-storage-cost/
Verified: 2026-10-25

**Quick answer:** 1 TB stored for a month costs $6.95 on Backblaze B2 and $10.24 on Bunny Storage in a single standard region. B2 also includes free egress equal to three times stored data; Bunny charges nothing to move data out of storage, but its CDN bills $0.01 per GB to deliver it.

- Backblaze B2 charges $6.95 per TB per month with the first 10 GB always free, so 1 TB stored costs $6.95 (checked 2026-10-25).
- Bunny Storage charges $0.01 per GB-month in a single standard region, so 1 TB stored costs $10.24 — about 32% more than B2 (calculated 2026-10-25).
- B2 includes free egress equal to three times average monthly stored data, so 1 TB stored permits 3 TB of downloads at no charge (checked 2026-10-25).
- Egress beyond three times storage costs $0.01 per GB on B2, so 1 TB stored plus 4 TB of downloads bills $17.19 (calculated 2026-10-25).
- Bunny Storage has no egress charge of its own, but delivery through Bunny's CDN bills $0.01 per GB in Europe and North America (checked 2026-10-25).
- B2 offers unlimited free egress to partner CDNs including Cloudflare, Fastly and bunny.net, and B2 Overdrive at $15 per TB-month removes the egress cap entirely (checked 2026-10-25).

Backblaze B2 and Bunny Storage both sell object storage by the month, and they disagree about almost everything else. One terabyte held for a month costs $6.95 on B2 and $10.24 on Bunny in a single standard region (calculated 2026-10-25). The gap is 32% on storage alone, and it flips depending on how much you download, because B2 bundles free egress at three times your stored volume while Bunny bundles none of the delivery. This page prices both under the same three scenarios so the storage rate stops being the whole story.

## What 1 TB costs on each

| Item | Backblaze B2 | Bunny Storage |
|---|---|---|
| Price per TB-month | $6.95 | $10.24 |
| Effective per GB-month | $0.0068 | $0.01 |
| Free allowance | First 10 GB always free | $1 monthly minimum |
| Regions | Multi-region, one rate | $0.02/GB for two, $0.025/GB for three |
| Higher tier | B2 Overdrive from $15/TB | SSD edge tier at $0.02/GB |
| Free egress | 3× average monthly storage | Storage to Bunny CDN only |
| Egress beyond that | $0.01 per GB | CDN rate, $0.01/GB EU and NA |

[[ad]]

## What the egress rules actually mean

Storage rates are the easy half. The bill is decided by how many times each stored byte leaves, and the two vendors answer that question differently.

| Monthly scenario | Backblaze B2 | Bunny Storage + CDN |
|---|---|---|
| 1 TB stored, no downloads | $6.95 | $10.24 |
| 1 TB stored, 1 TB downloaded | $6.95 | $20.48 |
| 1 TB stored, 4 TB downloaded | $17.19 | $51.20 |

B2 gives you egress equal to three times your average monthly stored data, so 1 TB stored permits 3 TB of downloads at no charge and the first scenario costs the same as the second (calculated 2026-10-25). Past that the rate is $0.01 per GB: the fourth terabyte adds $10.24, which makes the 4 TB month $17.19.

Bunny's model is the mirror image. Nothing is charged to move data from storage to the CDN, and there are no API fees, but the CDN that receives it bills $0.01 per GB in Europe and North America. Every delivered terabyte therefore costs $10.24 on top of storage, and the 4 TB month reaches $51.20 — three times B2's bill for the same workload (calculated 2026-10-25).

## How we measured it

We loaded Backblaze's cloud storage pricing page and both Bunny pricing pages on 2026-10-25 and used only the published rates: $6.95 per TB, the three-times egress rule and $0.01 per GB beyond it on one side; $0.01 per GB-month, the $1 minimum and the $0.01 per GB CDN rate on the other. The terabyte is read as 1,024 GB throughout. No bucket was created, nothing was uploaded and no transfer was measured.

## What changes in practice

Multi-region storage breaks the comparison in Bunny's favour at the low end and against it at the high end. One region is $0.01 per GB, two is $0.02 and three is $0.025, so spreading 1 TB across three regions costs $25.60 a month against B2's flat $6.95 — nearly four times as much for a feature B2 prices into a single rate.

The opposite trade appears if your CDN is already one of B2's partners. Backblaze lists Cloudflare, Fastly, bunny.net, CacheFly, CoreWeave, Equinix Metal, Vultr and phoenixNAP as destinations where egress is unlimited and free, which means a B2 bucket fronted by Bunny's CDN can pay $6.95 for storage and Bunny's $0.01 per GB for delivery, with no B2 egress bill at all. The 4 TB scenario becomes $6.95 plus $40.96, and the storage side of the bill stays at B2's rate.

Operational calls are the third line item. B2 makes Class A, B and C transactions free on pay-as-you-go and charges $0.004 per 10,000 Class D calls with the first 2,500 a day free; Bunny bundles API calls into the storage price. For a bucket serving millions of small objects, the two models diverge in ways a per-TB comparison hides.

## Bottom line

1 TB stored for a month costs $6.95 on B2 and $10.24 on Bunny, and the gap widens as downloads grow: $17.19 against $51.20 at 4 TB of egress. Store on B2 unless you need multi-region replication, and if you front it with a partner CDN, B2's egress cap stops mattering altogether.

**Sources:**
- Backblaze B2 pricing: https://www.backblaze.com/cloud-storage/pricing
- Bunny.net storage pricing: https://bunny.net/pricing/storage/
- Bunny.net CDN pricing: https://bunny.net/pricing/

---

# Bunny.net Pricing per TB in 2026: $10.24 for Europe and North America

URL: https://actualdatabase.com/bunny-net-pricing-per-tb/
Verified: 2026-10-25

**Quick answer:** 1 TB of CDN traffic from Europe or North America costs $10.24 on Bunny.net's standard network at $0.01 per GB. Asia and Oceania cost $30.72, South America $46.08 and the Middle East and Africa $61.44 for the same terabyte.

- Bunny.net charges $0.01 per GB for CDN traffic in Europe and North America, so 1 TB costs $10.24 (calculated 2026-10-25).
- Asia and Oceania are billed at $0.03 per GB, South America at $0.045 and the Middle East and Africa at $0.06 (checked 2026-10-25).
- The volume network charges a single global rate of $0.005 per GB for the first 500 TB, or $5.12 per TB (calculated 2026-10-25).
- Bunny.net applies a $1 monthly minimum, which buys 100 GB of European or North American traffic at $0.01 per GB (calculated 2026-10-25).
- Bunny Storage costs $0.01 per GB-month in a single standard region, so 1 TB stored is $10.24 a month with a $1 minimum (calculated 2026-10-25).
- The standard network runs on 119 points of presence while the volume network runs on 10 with a single global rate (checked 2026-10-25).

Bunny.net sells CDN traffic by the gigabyte with a rate that depends on where the bytes are delivered. In Europe and North America that rate is $0.01 per GB, so 1 TB costs $10.24; the same terabyte delivered from Asia, Oceania or Africa costs three to six times more (calculated 2026-10-25). The $1 monthly minimum is small enough to be irrelevant for anyone asking about terabytes, but the regional table is not. This page lays out the per-region rates, the volume network alternative and the storage line that usually sits next to the CDN bill.

## What 1 TB costs by region

| Region (standard network) | Per GB | Cost for 1 TB |
|---|---|---|
| Europe and North America | $0.01 | $10.24 |
| Asia and Oceania | $0.03 | $30.72 |
| South America | $0.045 | $46.08 |
| Middle East and Africa | $0.06 | $61.44 |
| Volume network, first 500 TB | $0.005 | $5.12 |

[[ad]]

## What the per-GB rate actually means

The rate is per region because the cost of moving a byte is per region. Bunny's standard network runs on 119 points of presence, and it bills the traffic that leaves each region at that region's price, so a global audience produces a blended rate rather than a single one. A site that serves 500 GB from Europe and 500 GB from Asia pays $5.12 plus $15.36, which is $20.48 for a terabyte that the headline rate would have priced at $10.24 (calculated 2026-10-25).

The volume network is the same product at a smaller footprint. It runs on 10 points of presence and charges one global rate of $0.005 per GB for the first 500 TB, with tiers at $0.004 per GB from 500 TB to 1 PB and $0.002 beyond that. For a workload that is not latency-sensitive in every country — large video files, backups, software updates — that is half the European rate for the same byte count.

The $1 minimum deserves a line because it is the one charge that does not scale. At $0.01 per GB it buys 100 GB, so any site serving less than 100 GB a month in Europe pays a flat $1.00. That is a rounding error at terabyte scale and a 100% surcharge at 50 GB.

## How we measured it

We loaded the CDN pricing page and the storage pricing page on 2026-10-25 and took the published per-GB rates, the region labels and the minimums from them. The terabyte is read as 1,024 GB and multiplied by the published rate, with the result dated. No file was uploaded, no traffic was served and no benchmark was run; the blended-rate example is arithmetic on the same table, not a measurement of a real traffic mix.

## What changes in practice

Storage is a second bill, and it is priced close to the CDN. Bunny Storage charges $0.01 per GB-month in a single standard region, with $0.005 per GB for each additional region up to nine, and the same $1 minimum. A terabyte stored in one region therefore costs the same $10.24 as a terabyte delivered from Europe — which means a small video site holding 1 TB and serving 1 TB pays about $20.48 in total (calculated 2026-10-25), split evenly between holding the bytes and moving them.

Two multipliers catch people out. Spreading storage across three regions costs $0.025 per GB instead of $0.01, so the same terabyte becomes $25.60 a month before any traffic. And the SSD edge tier doubles the storage rate to $0.02 per GB per region, which is $20.48 per TB, without changing the CDN rate at all.

The regional rate is the one to model before committing. A predominantly European audience is the cheapest case Bunny sells at $0.01 per GB, and a genuinely global one can land at two to three times the headline depending on the mix. Pulling the traffic mix from an existing CDN's logs before switching is the difference between a $10 and a $30 terabyte.

## Bottom line

For European and North American traffic, 1 TB on Bunny.net costs $10.24, and the $1 minimum only matters below 100 GB. Model the region mix before you commit: the same terabyte is $30.72 from Asia and $61.44 from Africa, and the volume network halves the European rate at $5.12 if 10 points of presence are enough.

**Sources:**
- Bunny.net CDN pricing: https://bunny.net/pricing/
- Bunny.net storage pricing: https://bunny.net/pricing/storage/

---

# What 1 TB of Video Traffic Costs in 2026: Bunny vs Fastly vs B2

URL: https://actualdatabase.com/cdn-cost-for-1tb-of-video-traffic/
Verified: 2026-10-25

**Quick answer:** 1 TB of video costs $10.24 on Bunny.net in Europe and North America, $110.88 on Fastly, and $6.95 on Backblaze B2 when the same terabyte is stored there, because B2 gives free egress up to three times stored data.

- 1 TB of video traffic costs $10.24 on Bunny.net's European and North American CDN at $0.01 per GB (calculated 2026-10-25).
- The same terabyte costs $110.88 on Fastly, where the first 100 GB is free and the remaining 924 GB bill at $0.12 per GB (calculated 2026-10-25).
- Backblaze B2 serves 1 TB of egress free when 1 TB is stored, because the allowance is three times average monthly storage, so the month bills $6.95 (calculated 2026-10-25).
- Storing only 100 GB on B2 drops the free allowance to 300 GB and turns a 1 TB month into $7.92 (calculated 2026-10-25).
- Video objects are large enough that request fees barely register: a terabyte of 10 MB files is about 100,000 requests, inside Fastly's free 1 million (calculated 2026-10-25).
- Bunny.net applies a $1 monthly minimum that does not bind at this volume, and Fastly publishes no CDN minimum at all (checked 2026-10-25).

Video is the workload that makes bandwidth pricing legible, because a terabyte of video is a terabyte of egress and nothing else. On the published rates captured on 2026-10-25, that terabyte costs $10.24 on Bunny.net from Europe or North America, $110.88 on Fastly, and $6.95 on Backblaze B2 if the same terabyte is stored there — a tenfold spread for the same bytes (calculated 2026-10-25). The interesting part is not the spread but the free allowances, which reorder the ranking depending on how much you store.

## What 1 TB of video costs on each

| Provider | Published rate | Free allowance | Cost for 1 TB |
|---|---|---|---|
| Bunny.net CDN | $0.01/GB, EU and NA | $1 monthly minimum | $10.24 |
| Fastly | $0.12/GB, EU and NA | 100 GB per month | $110.88 |
| Backblaze B2, 1 TB stored | $6.95/TB storage | 3 TB egress | $6.95 |
| Backblaze B2, 100 GB stored | $6.95/TB storage | 300 GB egress | $7.92 |

[[ad]]

## Where the free allowances change the answer

Fastly's 100 GB free is worth $12.00 off a European bill, which is real money at 200 GB and immaterial at 1 TB. It is also per region per month, so a video that only ships in North America gets the full credit there.

Backblaze's allowance is the one that changes the ranking. Egress is free up to three times your average monthly stored data, so storing 1 TB covers a 1 TB serving month completely and the bill is storage alone at $6.95. Store only 100 GB and the allowance falls to 300 GB, leaving 724 GB billable at $0.01 per GB — $7.24 of egress on top of about $0.68 of storage, or $7.92 (calculated 2026-10-25). Storing more is cheaper per served terabyte on B2, which is the opposite of how CDNs price.

Bunny has no egress allowance to change; its $1 minimum is consumed by the first 100 GB and never binds at terabyte scale. Its rate is flat per region, so the only variable is where the viewers are: the same terabyte costs $30.72 from Asia and $61.44 from Africa.

## How we measured it

We loaded the Bunny.net, Fastly and Backblaze B2 pricing pages on 2026-10-25 and took the printed rates, the free allowances and the minimums from them. Each terabyte is read as 1,024 GB, the free allowance is applied once, and the arithmetic is dated. No video was uploaded or served and no benchmark was run. The B2 scenarios assume the traffic is fetched directly from B2 rather than through a partner CDN, because through a partner the egress is unlimited but the partner's delivery rate applies instead.

## What changes in practice

Requests are not the problem with video, and that is the reason the comparison stays clean. A terabyte of 10 MB files is about 100,000 requests; even at $0.01 per 10,000 that is $1.00 on Fastly, and Bunny bundles its API and request handling into the storage and delivery rates. For an image library the same terabyte could be tens of millions of requests, and the request meter would overtake the bandwidth meter entirely.

The real decision is whether you need an edge at all. B2 at $6.95 is the cheapest number in the table, and it is a storage endpoint in a handful of regions, not a CDN — viewers far from those regions get higher latency and worse throughput, which for video means buffering rather than a slower page. Bunny at $10.24 is 47% more and buys 119 points of presence (calculated 2026-10-25). Fastly at $110.88 buys a premium network and the tooling around it.

There is a hybrid worth knowing about. B2 lists Cloudflare, Fastly, bunny.net, CacheFly and several compute providers as partners where egress is unlimited and free, so a B2 origin behind Bunny's CDN pays $6.95 for storage plus Bunny's $10.24 for delivery and no B2 egress at all. That combination costs more than either alone and is the right answer when the origin has to be object storage rather than a CDN's own storage.

## Bottom line

Serve 1 TB of video a month from Bunny.net and the bill is $10.24; from Fastly it is $110.88; from B2 direct it is $6.95 as long as you store at least a third of what you serve. The free allowance that matters is B2's three-times-stored rule, and it is the only one on this list that gets more generous the more you keep.

**Sources:**
- Bunny.net CDN pricing: https://bunny.net/pricing/
- Fastly pricing: https://www.fastly.com/pricing
- Backblaze B2 pricing: https://www.backblaze.com/cloud-storage/pricing

---

# The cheapest place to run a 512 MB app in 2026: $4 on DigitalOcean, $5 on Railway, $7 on Render

URL: https://actualdatabase.com/cheapest-place-to-run-a-512mb-app-2026/
Verified: 2026-10-25

**Quick answer:** A 512 MB always-on service costs $4.00 a month on a DigitalOcean basic Droplet, $5.00 on Railway's Hobby plan and $7.00 on Render's Starter instance. The Droplet is cheapest and includes 500 GiB of transfer, but it is an unmanaged Linux box, while the other two hand you git deploys and managed TLS.

- One 512 MB always-on service costs $4.00 a month on a DigitalOcean basic Droplet, $5.00 on Railway Hobby and $7.00 on Render Starter (checked 2026-10-25).
- The $4 Droplet includes 1 vCPU, 10 GiB of SSD and 500 GiB of outbound transfer, against 5 GB of bandwidth on Render's Hobby workspace (checked 2026-10-25).
- Railway's Hobby floor is $5.00 and includes $5 of usage, which covers 512 MB of memory for a month but leaves no room for CPU time (calculated 2026-10-25).
- At 5 GB of egress and 0.02 vCPU, Railway lands at about $5.65; at 0.1 vCPU the same service costs about $7.25 (calculated 2026-10-25).
- Render's free 512 MB instance costs $0 but spins down after 15 minutes without traffic and takes about a minute to wake, so always-on starts at $7.00 (checked 2026-10-25).
- The spread across the three options is $3.00 a month, or $36 a year, which is small next to the engineering time each one demands (calculated 2026-10-25).

One process, 512 MB of RAM, always on, light traffic: the smallest production shape there is. Priced on the three platforms people usually compare, it costs $4.00 a month on a DigitalOcean Droplet, $5.00 on Railway's Hobby plan and $7.00 on Render's Starter instance. The gap between the cheapest and the most expensive is $36 a year, which is small enough that the decision should be made on operations rather than on price — but the prices still need to be right, because two of the three are not what they look like.

## One small service, three invoices

| Option | Monthly | What the price includes | What it costs you |
|---|---|---|---|
| DigitalOcean basic Droplet | $4.00 | 512 MiB, 1 vCPU, 10 GiB SSD, 500 GiB transfer | you run the OS, proxy and deploys |
| Railway Hobby | $5.00 | 512 MB always on, git deploys, TLS | $5 minimum leaves no CPU budget |
| Railway at 0.02 vCPU | $5.65 | adds ~0.4 vCPU-hours a day plus 5 GB egress | CPU is metered separately |
| Railway at 0.1 vCPU | $7.25 | the same service, busier | cost scales with work done |
| Render Starter | $7.00 | 512 MB, 0.5 CPU, always on, 5 GB bandwidth | bandwidth overage at $0.15/GB |
| Render Free | $0.00 | 512 MB, sleeps after 15 minutes idle | ~1 minute cold start |

[[ad]]

## What each price actually buys

The Droplet's $4 is the only flat price in the table. It buys a virtual machine with 1 vCPU, 10 GiB of SSD and 500 GiB of outbound transfer, and it does not change if the app works harder that month. In exchange, nothing is managed: the OS image is minimal and yours to patch, TLS is yours to renew, and a deploy is whatever script you wrote. DigitalOcean bills per second with a 60-second minimum, so a Droplet created and destroyed an hour later costs a fraction of a cent.

Railway's $5 is a floor rather than a price. Hobby is a $5 minimum usage plan that includes $5 of usage, and 512 MB of memory alone costs about $5.00 over a 30-day month. That means the minimum and the memory bill cancel out exactly, leaving nothing for CPU or egress, so any real work pushes the total above $5. At 0.02 vCPU averaged over the month the bill is about $5.65; at 0.1 vCPU it is about $7.25 (calculated 2026-10-25).

Render's $7 is flat and predictable, and it is the third rung of a ladder that starts at free. The free instance has the same 512 MB but spins down after 15 minutes without inbound traffic and takes about a minute to wake, so it is not always-on. $7 buys the same memory with the sleeping removed, plus 5 GB of bandwidth before the $0.15 per GB overage starts.

## How we measured it

All three pricing pages were fetched on 2026-10-25: Render's for the $7 compute tier and its 5 GB Hobby allowance, Railway's for the $5 minimum and the per-second memory and CPU rates, and DigitalOcean's Droplet pricing for the $4 512 MiB tier and its 500 GiB of transfer. Railway's figures required arithmetic — memory and CPU seconds multiplied by 2,592,000 — and the vCPU averages are stated assumptions, not published numbers. No app was deployed on any of the three.

## What changes in practice

The database is bigger than the app on every platform. DigitalOcean's managed database starts at $15 a month, Render's Postgres at $6 for 256 MB and $19 for 1 GB, and Railway bills a database as just another service at the same rates, which for 256 MB plus a little CPU is roughly $4 to $6. A $4 app with a $15 database is a $19 project, and that reversal surprises people who compared only the compute line.

Railway's metering cuts both ways. A service that is genuinely idle most of the day is cheap in CPU terms and lands near its $5 floor, while a service that does real work pays for it — the same 512 MB costs $5.65 or $7.25 depending on how busy it is. Render charges $7 whether the CPU sits at 2 percent or 40 percent, which is worth more than the $1.35 difference if the load is spiky and hard to predict.

The Droplet's hidden cost is time, and it is not small. A $4 Droplet needs a firewall, a reverse proxy, certificate renewal and a deploy path before it serves anything; that is a day of work once and perhaps an hour a month afterwards. At any reasonable hourly rate the first year of a Droplet costs more than three years of Render's $7 plan, which is why the cheapest line item is rarely the cheapest choice.

## Bottom line

Pay $4 for the Droplet if you want a flat bill and are willing to own the server, $5 to $7.25 on Railway if the app's CPU use is low and variable, and $7 on Render if you want the same price every month with nothing to administer. For one 512 MB service with modest traffic, Render at $7 is the least surprising invoice and DigitalOcean at $4 is the cheapest by $36 a year (calculated 2026-10-25).

**Sources:**
- Render pricing: https://render.com/pricing
- Railway pricing: https://railway.com/pricing
- DigitalOcean Droplet pricing: https://www.digitalocean.com/pricing/droplets

---

# Cloudflare Free vs Pro in 2026: What $20 a Month Actually Adds

URL: https://actualdatabase.com/cloudflare-plans-free-vs-pro/
Verified: 2026-10-25

**Quick answer:** Cloudflare Pro costs $20 per month billed annually ($25 monthly) and adds 225 Cloudflare Rules instead of 70, lossless image optimization, and the OWASP Core and Cloudflare Managed WAF rulesets with 20 custom rules. Free already includes DNS, the CDN, Universal SSL and unmetered DDoS protection.

- Cloudflare Free costs $0 and Pro costs $20/month billed annually or $25/month billed monthly (checked 2026-10-25).
- Pro includes 225 Cloudflare Rules and 25,000 Bulk Redirects, against 70 rules and 10,000 Bulk Redirects on Free (checked 2026-10-25).
- Free includes unmetered DDoS protection, the CDN, Universal SSL, the Free Managed Ruleset and WAF, but not Lossless Image Optimization (checked 2026-10-25).
- Neither Free nor Pro includes PCI DSS 4.0 or an uptime SLA; both start at Business, $200/month billed annually (checked 2026-10-25).
- Workers on Free allows 100,000 requests per day at 10 ms of CPU each, about 3 million requests a month (calculated 2026-10-25).
- Paying annually saves $60 a year on Pro — $240 against $300 at the monthly rate (calculated 2026-10-25).

Cloudflare's Free plan is genuinely free and genuinely load-bearing: DNS, the CDN, Universal SSL and unmetered DDoS protection all sit on it at $0. Pro costs $20 per month billed annually, or $25 month to month, and the difference is mostly about rules and managed security rulesets rather than traffic (checked 2026-10-25). The question worth answering is not what Pro lists, but which of the gaps a small site will actually collide with. This page puts both plans in one table, then separates the limits that matter from the marketing ones.

## What each plan includes

| Item | Free | Pro |
|---|---|---|
| Monthly price | $0 | $20 annual, $25 monthly |
| Cloudflare Rules | 70 | 225 |
| Bulk Redirects | 10,000 | 25,000 |
| CDN, DNS, Universal SSL | Included | Included |
| DDoS protection | Unmetered | Unmetered |
| WAF | Free Managed Ruleset | OWASP Core + Cloudflare Managed + 20 custom rules |
| Lossless image optimization | Not included | Included |
| PCI DSS 4.0 | Not included | Not included |
| Uptime SLA | Not included | Not included |
| Workers | 100,000 requests/day, 10 ms CPU | Paid add-on: $0.30/million requests |

[[ad]]

## What the $20 actually buys

Three things change at Pro, and only one of them is about traffic. The rule ceiling moves from 70 to 225, which matters to sites that have accumulated redirects and cache rules over years rather than months. The WAF gains the OWASP Core Managed Ruleset, the Cloudflare Managed Ruleset and 20 custom rules, which is a real security upgrade over the Free Managed Ruleset. And image optimization becomes lossless, which is a bandwidth and Core Web Vitals argument rather than a protection one.

What does not change is the part most people assume a paid plan buys. The plans page lists PCI DSS 4.0 and the uptime SLA as Business-tier features, and Business is $200 per month billed annually — $2,400 a year, a 10× step from Pro (calculated 2026-10-25). API Shield, AI Security for Apps and China Network Access sit above that. If your reason for upgrading is a contractual SLA or card-data compliance, Pro does not solve it and the next price you will be quoted is $200.

## How we measured it

We loaded Cloudflare's plans page and the Pro plan page on 2026-10-25 and read the comparison rows directly, including the Workers rates that are printed inside the same table. Where a feature is listed as unavailable on both Free and Pro, we recorded the tier it appears on rather than describing the feature. No account was created and no traffic was proxied through either plan; these are published figures with our arithmetic attached.

## What changes in practice

For a static content site, Pro mostly buys headroom you will not use. Seventy rules covers a normal redirect map and a handful of cache rules; 100,000 Workers requests a day is roughly 3 million a month (calculated 2026-10-25), which is more than most small sites serve, and static asset requests do not touch the Workers quota at all.

The 10 ms CPU ceiling on Free is the quieter limit. It is enough to rewrite a header or serve a cached JSON response, and not enough to render a page, resize an image on the fly or call two upstream APIs in sequence. A small site that starts doing real work at the edge will find the paid Workers line item long before it finds the 70-rule ceiling — and that is billed separately from Pro, at $0.30 per million requests beyond the included 10 million.

The one upgrade that pays for itself in a week is image optimization, and only if you serve images at scale. Lossless optimization on a media-heavy site cuts bytes off every image request, which shows up as bandwidth and as Largest Contentful Paint. On a site serving 20 images per page view, $20 buys less than a coffee per thousand page views; on a five-page brochure site it buys nothing measurable.

## Bottom line

Pro is $240 a year billed annually, or $300 at the monthly rate, and at that price it is worth it for the 225-rule ceiling, the managed WAF rulesets or lossless images — pick one. If what you need is an uptime SLA or PCI DSS 4.0, budget $2,400 a year for Business instead, because $20 a month does not buy either.

**Sources:**
- Cloudflare plans: https://www.cloudflare.com/plans/
- Cloudflare Pro plan: https://www.cloudflare.com/plans/pro/

---

# Cloudflare R2 Pricing Explained: 1 TB Stored and 10 Million Reads

URL: https://actualdatabase.com/cloudflare-r2-storage-pricing-explained/
Verified: 2026-10-25

**Quick answer:** 1 TB stored in Cloudflare R2 for a month costs $15.21 after the 10 GB free allowance, and 10 million Class B reads cost $0.00 because they sit exactly inside the free tier. Egress is free, so the total bill is $15.21.

- R2 Standard storage costs $0.015 per GB-month, with Class A operations at $4.50 per million and Class B at $0.36 per million (checked 2026-10-25).
- Egress from R2 is free, with no bandwidth charge for data transferred out (checked 2026-10-25).
- The free tier covers 10 GB-month of storage, 1 million Class A and 10 million Class B operations per month (checked 2026-10-25).
- 1 TB stored for a month costs $15.21 after the free allowance: 1,014 GB-months at $0.015 (calculated 2026-10-25).
- 10 million Class B reads sit exactly inside the free allowance, so 1 TB stored plus 10 million reads bills $15.21 (calculated 2026-10-25).
- Infrequent Access storage is $0.01 per GB-month with Class B at $0.90 per million and no free tier, beating Standard only below roughly 5.5 million reads a month at 1 TB (calculated 2026-10-25).

Cloudflare R2 prices storage by the GB-month, operations by the million, and egress at zero. One terabyte held for a month costs $15.21 after the 10 GB free allowance, and 10 million Class B reads cost nothing at all because the free tier stops exactly there (calculated 2026-10-25). That combination is unusual enough that the arithmetic is worth showing line by line, because the number that decides the bill is not the storage rate — it is which operations you are actually doing.

## What each R2 line item costs

| Component | Rate | Free tier | Bill for 1 TB + 10M reads |
|---|---|---|---|
| Storage (Standard) | $0.015 per GB-month | 10 GB-month | $15.21 |
| Class A operations | $4.50 per million | 1 million/month | $0.00 |
| Class B operations | $0.36 per million | 10 million/month | $0.00 |
| Egress | $0.00 per GB | Unlimited | $0.00 |
| Total | — | — | $15.21 |

[[ad]]

## What the storage rate actually means

The storage charge is the only line that lands at this volume. Reading 1 TB as 1,024 GB, subtracting the 10 GB free allowance leaves 1,014 GB-months at $0.015, which is $15.21 (calculated 2026-10-25). Cloudflare's own worked example uses a round 1,000 GB and lands on $14.85; the 36¢ difference is the terabyte definition, not a rate difference, and it is the kind of thing that makes two quotes disagree.

Class B operations are reads, and the free 10 million covers a surprising amount of serving. A site with 1 TB of media and 10 million object reads a month pays nothing for access; reads only start to cost once they pass 10 million, at $0.36 per additional million. Class A operations — writes, lists, copies — are the expensive ones at $4.50 per million, with a 1 million monthly allowance that a backup job or a migration can exhaust in an afternoon.

Egress is free and unbounded, which is the line that makes R2 worth considering at all. There is no bandwidth component in the bill above, and there is no threshold after which bandwidth starts.

## How we measured it

We loaded the R2 pricing documentation and the R2 product page on 2026-10-25 and took both storage and operation rates from them, then applied the free tier before the paid rate. The 1 TB figure is read as 1,024 GB and the free allowance is applied once, as a monthly credit. No bucket was created, no data was uploaded and no benchmark was run; the read volume of 10 million is the scenario, not a measurement.

## What changes in practice

Infrequent Access is the trap for anyone who reads as far as the storage rate. It is $0.01 per GB-month against $0.015, which looks like a third off, but Class B operations cost $0.90 per million instead of $0.36 and the free tier does not apply to it. At 1 TB stored with 10 million reads, Standard bills $15.21 and Infrequent Access bills $19.24 before any retrieval charge, because you pay for all 10 million reads at the higher rate (calculated 2026-10-25). IA only wins below about 5.5 million reads a month at that volume, and data retrieval at $0.01 per GB adds another variable on top.

Rounding matters more than the rate at small scale. R2 bills the next whole unit: 1.1 GB-months becomes 2, and 1,000,001 operations become 2 million. A project storing 11 GB and doing 1,000,001 Class A operations pays for 2 GB and 2 million operations, which is more than double the free allowance it thought it had.

The last thing to plan for is the Class A count, not the storage count. Serving objects is cheap; rewriting them is not. A build that re-uploads 100,000 files a day runs 3 million Class A operations a month, which is $9.00 after the free million at $4.50 per million — more than half the cost of the terabyte it is rewriting (calculated 2026-10-25).

## Bottom line

At $15.21 for 1 TB stored and 10 million reads, R2's bill is dominated by the storage rate and nothing else, which is the point: the bandwidth charge that usually dominates object-storage invoices is $0 here. Stay on Standard unless a terabyte is read fewer than 5.5 million times a month, and watch the write count rather than the read count.

**Sources:**
- Cloudflare R2 pricing: https://developers.cloudflare.com/r2/pricing/
- Cloudflare R2 product page: https://www.cloudflare.com/developer-platform/products/r2/

---

# Cloudflare Workers Pricing: Free vs Paid, and What 10 Million Requests Cost

URL: https://actualdatabase.com/cloudflare-workers-pricing-free-vs-paid/
Verified: 2026-10-25

**Quick answer:** Workers Free allows 100,000 requests a day with 10 ms of CPU each, roughly 3 million requests a month. Workers Paid has a $5.00 monthly minimum covering 10 million requests and 30 million CPU-ms, so 10 million requests at 7 ms of CPU costs $5.80.

- Workers Free allows 100,000 requests a day at 10 ms of CPU per invocation, about 3 million requests a month (calculated 2026-10-25).
- Workers Paid has a $5.00 monthly minimum per account that includes 10 million requests and 30 million CPU-ms (checked 2026-10-25).
- Overage on Paid is $0.30 per million requests and $0.02 per million CPU-ms (checked 2026-10-25).
- 10 million requests a month costs $5.80 on Paid at 7 ms of CPU per request: $0.00 in request overage and $0.80 in CPU overage (calculated 2026-10-25).
- 100 million requests at 7 ms costs $45.40 on Cloudflare's own worked example (checked 2026-10-25).
- There is no charge for egress or bandwidth, and requests for static assets are free and unlimited on both plans (checked 2026-10-25).

Cloudflare Workers is priced on two meters, requests and CPU time, and the free plan gives you a generous amount of the first and a tight amount of the second. Free allows 100,000 requests a day at 10 ms of CPU per invocation; Paid has a $5.00 monthly minimum that includes 10 million requests and 30 million CPU-ms. At 7 ms of CPU per request, 10 million requests a month costs $5.80 (calculated 2026-10-25). This page shows where the meters trip and what each volume band actually bills.

## What each plan includes

| Item | Free | Paid |
|---|---|---|
| Monthly minimum | $0 | $5.00 per account |
| Requests included | 100,000 per day | 10 million per month |
| Request overage | Not available | $0.30 per million |
| CPU per invocation | 10 ms | 30 s default, 5 min maximum |
| CPU included | No charge for duration | 30 million CPU-ms per month |
| CPU overage | Not available | $0.02 per million CPU-ms |
| Egress | Free | Free |
| Static asset requests | Free, unlimited | Free, unlimited |

[[ad]]

## What the meters actually charge

Request volume is the meter people watch and CPU time is the meter that decides the bill. A worker that answers from a cache in 2 ms costs almost nothing in CPU and almost everything in requests; a worker that fetches three APIs and renders HTML costs the reverse.

| Requests per month | Request overage | CPU-ms at 7 ms | Monthly bill |
|---|---|---|---|
| 1 million | $0.00 | 7 million (inside 30M) | $5.00 |
| 5 million | $0.00 | 5 million over | $5.10 |
| 10 million | $0.00 | 40 million over | $5.80 |
| 50 million | $12.00 | 320 million over | $23.40 |
| 100 million | $27.00 | 670 million over | $45.40 |

The 100 million row is Cloudflare's own worked example, which is useful because it removes any argument about method: $5.00 minimum, $27.00 of request overage at $0.30 per million beyond the included 10 million, and $13.40 of CPU overage at $0.02 per million CPU-ms. The same page shows a caching variant — 100 million requests with an 80% cache hit rate, so only 20 million invoke the worker — costing $34.20, because the request charge stays the same and the CPU charge collapses.

## How we measured it

We loaded the Workers pricing documentation and the Workers limits page on 2026-10-25, took the published allowances and rates, and applied a 7 ms CPU figure per request because that is the number Cloudflare uses in its own examples. No worker was deployed and no traffic was generated; these are published rates multiplied out, with the arithmetic dated. The 7 ms assumption is the one variable that changes the CPU column, and it is stated rather than buried.

## What changes in practice

The 10 ms CPU ceiling on Free is the real upgrade trigger, not the 100,000 requests. Three million requests a month is more than most small sites serve, but 10 ms is not enough to resize an image, parse a large JSON payload and re-serialise it, or talk to two services in sequence. The moment a worker needs 20 ms, Paid is the only option and the bill starts at $5.

Cache hit rate is the lever that keeps the bill flat. Requests are charged whether or not they reach your code, so a worker sitting behind a cache still pays $0.30 per million for the privilege of being skipped. Cloudflare's 80% example is the proof: 100 million requests cost $45.40 cold and $34.20 warm, and the $11.20 difference is entirely CPU time that never happened.

There is no bandwidth line at all, which is why the plan suits media-heavy sites. Static asset requests are free and unlimited, egress is free, and subrequests a worker makes downstream are not billed. A site serving 1 TB of images through Workers pays for invocations only.

## Bottom line

Below roughly 3 million requests a month and 10 ms of CPU per request, Free costs $0 and does the job. Above that, budget the $5.00 minimum plus $0.30 per million requests and $0.02 per million CPU-ms: 10 million requests lands at $5.80, and the bill only becomes a real line item past 50 million requests, where it reaches $23.40.

**Sources:**
- Cloudflare Workers pricing: https://developers.cloudflare.com/workers/platform/pricing/
- Cloudflare Workers limits: https://developers.cloudflare.com/workers/platform/limits/

---

# Transactional email: cost per 1,000 emails at 1,000, 10,000 and 50,000 a month

URL: https://actualdatabase.com/cost-per-1000-transactional-emails/
Verified: 2026-10-25

**Quick answer:** At 10,000 emails a month Postmark and Mailgun both cost $15.00, which is $1.50 per 1,000, against $20.00 on Resend. At 50,000 emails Resend's $20 Pro plan is the cheapest at $0.40 per 1,000, ahead of Mailgun's $35.00 Foundation tier and Postmark's $66.00 Platform tier.

- Resend Free covers 3,000 emails a month with a 100-email daily cap, and Resend Pro is $20/month for 50,000 emails (checked 2026-10-25).
- Postmark's cheapest paid plan is $15/month for 10,000 emails, and Postmark states it sells no tier between 100 emails a month and 10,000 (checked 2026-10-25).
- Mailgun Free allows 100 emails a day, Basic is $15/month for 10,000 and Foundation is $35/month for 50,000 (checked 2026-10-25).
- 10,000 emails a month costs $15.00 on Postmark Basic and Mailgun Basic, or $1.50 per 1,000, against $20.00 on Resend Pro (calculated 2026-10-25).
- 50,000 emails a month costs $20.00 on Resend Pro ($0.40 per 1,000), $35.00 on Mailgun Foundation ($0.70) and $66.00 on Postmark Platform ($1.32) (calculated 2026-10-25).
- 1,000 emails a month costs $0 on Resend and Mailgun inside their 100-a-day free caps, and $15.00 on Postmark's cheapest paid tier (calculated 2026-10-25).

Transactional email is priced in tiers, and the tier boundaries matter more than the headline rate: the cheapest provider at 10,000 emails is not the cheapest at 50,000, because one of them stops charging at 50,000 and the others keep metering.

At 10,000 emails a month Postmark and Mailgun both cost $15.00 — $1.50 per 1,000 — and Resend costs $20.00. At 50,000 the order flips: Resend's $20 covers the whole volume at $0.40 per 1,000, against Mailgun's $35.00 and Postmark's $66.00.

## What each volume costs

| Monthly volume | Resend | Postmark | Mailgun |
|---|---|---|---|
| 1,000 | $0 inside 100/day cap | $15.00 Basic | $0 inside 100/day cap |
| 10,000 | $20.00 Pro | $15.00 Basic | $15.00 Basic |
| 50,000 | $20.00 Pro | $66.00 Platform | $35.00 Foundation |

Postmark's figures use its cheapest plan at each volume: Basic at 10,000 and Platform at 50,000, since Platform's $1.20 per-1,000 overage beats Basic's $1.80. Mailgun's use Basic then Foundation, whose 50,000 included emails cover the volume exactly.

[[ad]]

## What the numbers actually mean

Resend has the strangest curve of the three. Free carries 3,000 emails a month with a 100-a-day cap, and the next step is $20 for 50,000. That means a sender at 5,000 emails a month pays $20 and a sender at 45,000 pays the same $20, so the effective rate falls from $4.00 per 1,000 to $0.44 as volume grows. Past 50,000 the curve flattens back into a metered line: another 50,000 emails cost $45.00 in overage at $0.90 per 1,000.

Postmark's curve is inverted: cheap at the bottom, expensive at the top. Basic is $15 for 10,000 and $1.80 per 1,000 beyond, which is $87.00 at 50,000 before you switch to Platform at $66.00. Both are the same allowance at different marginal rates, so the choice between them is purely a question of how much overage you expect.

Mailgun sits between the two. Basic matches Postmark at $15 and $1.80, but the next tier is a bigger bucket rather than a cheaper marginal rate: Foundation at $35 for 50,000. At exactly 50,000 emails that is $0.70 per 1,000, cheaper than either Postmark tier.

The free tiers are where the per-1,000 number collapses to zero, and they are not equivalent. Resend and Mailgun both meter 100 emails a day, which allows about 3,000 a month. Postmark's free plan is 100 emails a month and forbids overages, so 1,000 emails a month is a $15 bill there and $0 on the other two.

## How we measured it

Prices, allowances and overage rates are quoted from the Resend, Postmark and Mailgun pricing pages, all opened on 2026-10-25. Nothing was sent and no deliverability test was run. Each total is the cheapest published plan that covers the volume, plus overage at that plan's published per-1,000 rate where the volume exceeds the allowance; those calculations are marked as such, and the effective per-1,000 figures divide the total by the volume.

## What changes in practice

The plan boundary is the risk, not the rate. Resend bills overage only when pay-as-you-go is enabled under Transactional Overages, so a product that crosses 50,000 emails in a launch week stops sending rather than paying $0.90 per 1,000. Postmark charges overage automatically and does not roll unused emails over, so a quiet month does not buy credit for a loud one.

Minimum monthly spend is the second consideration. Postmark's floor is $15 whatever you send, and Resend's is $20 once you exceed 3,000 emails. For a product doing 2,000 transactional emails a month, that is a $180 to $240 annual commitment to a service whose marginal cost is close to nothing.

The third is what the per-1,000 rate excludes. Dedicated IPs, inbound processing, longer retention and higher sending limits all sit on higher tiers, and Postmark gates dedicated IPs at 300,000 emails a month while Resend gates them behind its $90 Scale plan plus a $30 add-on.

## Bottom line

Under 10,000 emails a month, Postmark or Mailgun at $15.00 beats Resend's $20.00, and the free daily caps at Resend and Mailgun mean small senders can start at $0. From 20,000 emails upward, Resend Pro's flat $20 is the cheapest published option until 50,000, after which its $0.90 per 1,000 keeps it competitive against Mailgun's $1.30 and Postmark's $1.20.

**Sources:**
- Resend pricing: https://resend.com/pricing
- Postmark pricing: https://postmarkapp.com/pricing
- Mailgun pricing: https://www.mailgun.com/pricing/

---

# Datadog for 10 Hosts: Per-Host Price and the Monthly Bill

URL: https://actualdatabase.com/datadog-cost-for-10-hosts/
Verified: 2026-10-25

**Quick answer:** Datadog Infrastructure Pro is $15 per host per month billed annually, or $18 on-demand, so ten hosts cost $150 per month or $1,800 per year on annual billing, and $180 per month or $2,160 per year on on-demand pricing. Infrastructure Enterprise is $23 per host on annual billing and $27 on-demand, which puts ten hosts at $2,760 a year or $3,240.

- Datadog Infrastructure Pro is $15 per host per month billed annually, or $18 per host on on-demand billing (checked 2026-10-25).
- Ten hosts on Infrastructure Pro cost $150 per month, or $1,800 per year, on annual billing (calculated 2026-10-25).
- The same ten hosts cost $180 per month, or $2,160 per year, on on-demand billing (calculated 2026-10-25).
- Infrastructure Enterprise is $23 per host per month annually and $27 on-demand, so ten hosts cost $2,760 or $3,240 a year (checked 2026-10-25).
- The free plan covers up to 5 hosts and is a separate plan, so ten hosts means ten paid host units rather than five free ones (checked 2026-10-25).
- Adding APM at $31 per host per month raises the ten-host bill to $460 per month, or $5,520 per year (calculated 2026-10-25).

Datadog's infrastructure pricing is one number per host, and the whole question is which of two numbers you get. Infrastructure Pro is $15 per host per month billed annually, or $18 per host on on-demand billing. Ten hosts therefore cost $150 per month and $1,800 per year on annual billing, or $180 per month and $2,160 per year on on-demand. Enterprise moves to $23 per host annually and $27 on-demand, which is $2,760 or $3,240 a year for the same ten hosts. The free plan does not discount the first five: it stops at five hosts entirely.

## Per-host infrastructure prices

| Plan | Annual billing, per host/month | On-demand, per host/month | Ten hosts, annual billing |
| --- | --- | --- | --- |
| Infrastructure Pro | $15 | $18 | $150 per month, $1,800 per year |
| Infrastructure Enterprise | $23 | $27 | $230 per month, $2,760 per year |

[[ad]]

## What ten hosts bill as you add products

| Product mix, ten hosts | Annual rate per host/month | Monthly bill | Annual bill |
| --- | --- | --- | --- |
| Infrastructure Pro only | $15 | $150 | $1,800 |
| Pro plus APM | $46 ($15 + $31) | $460 | $5,520 |
| Pro plus APM plus network monitoring | $51 | $510 | $6,120 |

Add-on rates are the published annual per-host prices for APM at $31 and network monitoring at $5 (calculated 2026-10-25).

## What the numbers actually mean

The base infrastructure bill is the smallest part of a Datadog invoice once instrumentation spreads. At $15 per host, ten hosts cost $1,800 a year, but the same ten hosts with APM attached cost $5,520, and adding network monitoring takes it to $6,120 (calculated 2026-10-25). APM alone is twice the price of infrastructure monitoring, which is why host counts quoted in isolation mislead: a ten-host estate with three products on it is a $6,000-a-year account, not an $1,800 one.

Annual billing is worth 16.7% on Pro and 14.8% on Enterprise (calculated 2026-10-25). On ten Pro hosts that is $360 a year, and the trade is a twelve-month commitment against a set of hosts that may not exist in twelve months. For a stable fleet the discount is free money; for an autoscaling group that swings between eight and fifteen hosts, the commitment can cost more than it saves.

Five hosts is a hard cliff rather than a ramp. Datadog's free plan covers up to 5 hosts with 1-day metric retention, and there is no published partial-credit structure that makes hosts six through ten cheaper. A ten-host account pays ten times the per-host rate, full stop, so consolidating workloads onto fewer, larger hosts is a real cost lever rather than a rounding exercise.

The on-demand rate exists for a reason. At $18 per host there is no annual commitment and no discount, which is the correct choice for a short-lived environment: ten on-demand hosts run for eight weeks cost about $360, against $1,800 committed for a full year (calculated 2026-10-25).

## How we measured it

Prices come from the Datadog pricing page and price list fetched on 2026-10-25, where the per-host figures appear as starting-at rates with the billed-annually or on-demand qualifier, and the same numbers are carried in the embedded price data that drives the page's billing toggle. No agents were installed, no hosts were onboarded and no invoice was produced. Ten-host totals, add-on sums and percentages are arithmetic on those published rates and are marked calculated.

## What changes in practice

Price the products, not the hosts. Infrastructure is $15 per host; APM is $31. A budget built on host count will be wrong by a factor of two or three once tracing and network monitoring are enabled, and the published per-host add-on rates are the only way to see that before the invoice.

Count hosts deliberately. Because ten hosts cost exactly ten units with no volume break at this size, merging small instances or dropping idle staging hosts is the cheapest optimization available, worth $15 per host per month on Pro.

Choose the billing term against fleet stability. Annual billing saves $360 a year on ten Pro hosts, but only if all ten run for twelve months; a fleet that shrinks by three hosts three months in gives most of that back.

## Bottom line

Ten hosts on Datadog Infrastructure Pro is $150 per month on annual billing, $1,800 a year, or $180 per month on-demand at $2,160 a year (calculated 2026-10-25). Budget at least $460 a month if APM is part of the plan, because at $31 per host the tracing line is the one that changes the size of the bill.

**Sources:**
- Datadog pricing: https://www.datadoghq.com/pricing/
- Datadog price list: https://www.datadoghq.com/pricing/list/

---

# Datadog Log Retention Cost: 15-Day vs 30-Day Pricing at 100 GB a Month

URL: https://actualdatabase.com/datadog-log-retention-cost/
Verified: 2026-10-25

**Quick answer:** Datadog charges $0.10 per ingested GB and then bills indexing per million events with a retention-based rate: $1.70 per million events per month at 15-day retention and $2.50 at 30-day retention on annual billing, or $2.55 and $3.75 on-demand. A hundred gigabytes a month costs $10 to ingest, and the annual bill lands between $120 and about $3,000 depending on how many events those gigabytes contain.

- Datadog log ingest is $0.10 per GB per month, the same on annual and on-demand billing (checked 2026-10-25).
- Indexing at 15-day retention is $1.70 per million log events per month on annual billing, or $2.55 on-demand (checked 2026-10-25).
- Indexing at 30-day retention is $2.50 per million log events per month on annual billing, or $3.75 on-demand (checked 2026-10-25).
- Doubling retention from 15 to 30 days costs 47% more per million events on annual billing (calculated 2026-10-25).
- 100 GB a month of log ingest costs $10 per month, or $120 per year, before indexing (calculated 2026-10-25).
- 100 GB containing 10 million events bills $27 per month at 15-day retention and $35 at 30-day retention, or $324 and $420 per year (calculated 2026-10-25).

Datadog splits a log bill into two lines that are priced in different units, and the retention window only moves the second one. Ingest is $0.10 per GB. Indexing, the part that makes logs searchable, is charged per million events per month at a rate that depends on how long you keep them: $1.70 at 15-day retention and $2.50 at 30-day retention on annual billing, or $2.55 and $3.75 on-demand. A hundred gigabytes a month costs $10 to bring in, and what happens after that depends entirely on how many events those gigabytes hold.

## Retention rates per million log events per month

| Retention window | Annual billing | Month-to-month | On-demand |
| --- | --- | --- | --- |
| 3 days | $1.06 | $1.27 | $1.59 |
| 7 days | $1.27 | $1.52 | $1.91 |
| 15 days | $1.70 | $2.04 | $2.55 |
| 30 days | $2.50 | $3.00 | $3.75 |

Log ingest is a separate line at $0.10 per GB on every billing type (checked 2026-10-25).

[[ad]]

## What 100 GB a month bills at each retention window

| Events inside 100 GB | 15-day indexing | 30-day indexing | Total monthly, 15-day | Total monthly, 30-day |
| --- | --- | --- | --- | --- |
| 10 million | $17 | $25 | $27 | $35 |
| 50 million | $85 | $125 | $95 | $135 |
| 100 million | $170 | $250 | $180 | $260 |

Totals add the $10 monthly ingest charge to the indexing line (calculated 2026-10-25).

## What the numbers actually mean

Retention is a per-event decision, so its cost depends on log volume rather than log size. Doubling the window from 15 to 30 days raises the per-million rate from $1.70 to $2.50, which is 47% more for twice the searchable history (calculated 2026-10-25). That is a mild step compared with the four-tier ladder as a whole: 3-day retention at $1.06 is 38% below the 15-day rate and 58% below the 30-day rate.

The per-GB number is the least important figure on the page. At $0.10 per GB, a hundred gigabytes a month is $10, which is smaller than any realistic indexing line: the same hundred gigabytes holding 10 million events costs $17 a month to index at 15 days and $25 at 30 days (calculated 2026-10-25). Teams that budget from the ingest rate underestimate the bill by an order of magnitude, because ingest is priced per gigabyte and indexing is priced per event.

Event size is the hidden variable. Datadog publishes no bytes-to-events conversion, so two accounts shipping the same 100 GB can differ by a factor of ten in cost depending on whether that volume is verbose debug lines or compact JSON. The only reliable way to forecast is to count events in a sample of the actual stream, then multiply by the retention rate. A 100 GB stream of 1 KB events is roughly 100 million events and bills $180 a month at 15-day retention (calculated 2026-10-25); the same 100 GB in 10 KB events is 10 million and bills $27.

Flex Storage changes the shape of long retention. Storing logs at $0.05 per million events per month, with a one-month minimum, is 50 times cheaper per million events than 30-day indexing at $2.50 (calculated 2026-10-25), and it is designed for logs you need to keep but rarely query. The catch is that querying stored logs requires rehydration, and the page prices rehydration as scans at $0.10 per compressed GB.

## How we measured it

Rates come from the Datadog pricing page and price list fetched on 2026-10-25, where the Standard Indexing card lists retention options of 3, 7, 15 and 30 days and the per-million-event rates appear in the page's embedded price data with annual, month-to-month and on-demand variants. No logs were ingested, no index was created and no trial was started. Event totals, totals including ingest and percentage differences are arithmetic on published rates and are labelled calculated.

## What changes in practice

Decide retention per index, not per account. Because the 3-day rate is $1.06 against $2.50 for 30 days, routing debug logs to a short window and keeping only audit and error logs at 30 days is worth 58% on the events that move (calculated 2026-10-25).

Sample before you ship. Ingest at $0.10 per GB is cheap enough that volume feels free, but indexing multiplies by event count, so dropping noisy debug lines at the agent is the highest-leverage saving in the whole pipeline.

Use Flex Storage for compliance tails. At $0.05 per million events stored, a month of 10 million archived events costs $0.50 against $25 to index them for 30 days (calculated 2026-10-25), provided you accept rehydration before you can query them.

## Bottom line

A hundred gigabytes a month costs $10 to ingest and between $27 and $260 a month in total depending on event count and retention (calculated 2026-10-25). Set 15-day retention unless a query needs otherwise: the 30-day window costs 47% more per event and rarely earns it back.

**Sources:**
- Datadog pricing: https://www.datadoghq.com/pricing/
- Datadog price list: https://www.datadoghq.com/pricing/list/

---

# DigitalOcean Droplet prices in 2026: $4 to $96 a month, 500 GiB included

URL: https://actualdatabase.com/digitalocean-droplets-price-per-month/
Verified: 2026-10-25

**Quick answer:** DigitalOcean's basic Droplets start at $4 a month for 512 MiB of RAM, 1 vCPU, 10 GiB of SSD and 500 GiB of transfer, and run to $96 a month for 16 GiB of RAM and 8 vCPU. Every Droplet includes outbound transfer starting at 500 GiB a month, with overage at $0.01 per GiB and inbound traffic free.

- DigitalOcean's cheapest basic Droplet is $4 a month for 512 MiB of RAM, 1 vCPU, 10 GiB of SSD and 500 GiB of outbound transfer (checked 2026-10-25).
- The basic ladder runs $4, $6, $12, $18, $24, $48 and $96 a month, covering 512 MiB up to 16 GiB of RAM (checked 2026-10-25).
- Included transfer rises with each tier from 500 GiB to 6,000 GiB, and overage costs $0.01 per GiB with inbound traffic free (checked 2026-10-25).
- Droplets bill per second with a 60-second minimum or $0.01, whichever is greater, from January 1, 2026 (checked 2026-10-25).
- The $24 tier buys 2 vCPU, 4 GiB of RAM, 80 GiB of SSD and 4,000 GiB of transfer, doubling both RAM and transfer against the $12 tier (checked 2026-10-25).
- Optional backups add 20 percent of the Droplet price for weekly or 30 percent for daily, taking the $6 Droplet to $7.20 or $7.80 (calculated 2026-10-25).

DigitalOcean's Droplet ladder is one of the few hosting price lists that still reads like a price list: seven basic sizes, $4 to $96 a month, each with a fixed amount of RAM, vCPU, SSD and outbound transfer. The cheapest is $4 for 512 MiB, 1 vCPU, 10 GiB of disk and 500 GiB of transfer; the most expensive basic size is $96 for 16 GiB, 8 vCPU, 320 GiB of disk and 6,000 GiB of transfer. Because inbound traffic is free and overage is $0.01 per GiB, the bandwidth line rarely decides the tier — the RAM line does.

## The basic Droplet ladder

| RAM | vCPU | SSD | Transfer | Per month |
|---|---|---|---|---|
| 512 MiB | 1 | 10 GiB | 500 GiB | $4 |
| 1 GiB | 1 | 25 GiB | 1,000 GiB | $6 |
| 2 GiB | 1 | 50 GiB | 2,000 GiB | $12 |
| 2 GiB | 2 | 60 GiB | 3,000 GiB | $18 |
| 4 GiB | 2 | 80 GiB | 4,000 GiB | $24 |
| 8 GiB | 4 | 160 GiB | 5,000 GiB | $48 |
| 16 GiB | 8 | 320 GiB | 6,000 GiB | $96 |

Two jumps in that table are worth naming. Going from $12 to $24 doubles RAM from 2 GiB to 4 GiB, doubles transfer from 2,000 to 4,000 GiB and adds a second vCPU for exactly twice the money, which makes $24 the best value point in the ladder. Going from $18 to $24 buys the same 2 vCPU but doubles RAM, so anyone tempted by $18 for the extra CPU should look at $24 instead.

[[ad]]

## What the hourly rates actually mean

Every tier carries an hourly price as well, and those are not decoration. The $4 Droplet is $0.00595 an hour, the $24 tier is $0.03571 an hour, and billing is per second with a 60-second minimum or $0.01 from January 1, 2026. A Droplet created for a two-hour migration costs about a cent of compute at the low end, which is why temporary resize-and-destroy workflows are genuinely cheap on this platform.

The 60-second minimum is the only billing trap in the list. A script that creates and destroys twelve Droplets in a loop is billed 60 seconds each time rather than the two seconds each one lived, so automation that churns instances pays a floor per instance. For humans doing one resize a month, it never shows up.

The ladder stops at $96 for basic sizes, but the page lists CPU-Optimized Droplets above it, starting at $0.0625 an hour for 4 GiB and 2 vCPU. That is about $45.63 a month against $24 for the basic 4 GiB tier, and it is the tier to pick when the CPU itself is the bottleneck rather than the RAM (calculated 2026-10-25).

## How we measured it

Both DigitalOcean pages were fetched on 2026-10-25: the pricing hub for the $4 starting price, the transfer overage rate and the backup percentages, and the Droplets page for the full ladder with hourly and monthly prices. No Droplet was created and no invoice was generated. The backup figures and the 1,024 GiB overage example are arithmetic on published rates and are labelled as calculated.

## What changes in practice

Transfer is bundled generously enough that it hides problems. 500 GiB at $4 is 125 times the traffic a small app uses, and the $0.01 per GiB overage is a tenth of what the platform-as-a-service vendors charge, so a Droplet is almost never the expensive choice for bandwidth-heavy work. The decision is made by RAM and by how much of the machine you are willing to administer.

The Droplet price does not include the operations. There is no managed TLS, no git-based deploys and no zero-downtime release process unless you build one, and a $4 server that needs four hours of setup a month is not a $4 server. The honest comparison is against a $7 to $25 platform plan that does that work for you.

Add-ons move the total more than people expect. Backups at 20 or 30 percent are proportional to the Droplet, so a $96 server with daily backups is $124.80, and a managed database starts at $15 a month if you do not want Postgres on the same 512 MiB of RAM.

## Bottom line

Start at $4 if the app fits in 512 MiB and you are comfortable with a shell, and jump straight to $24 rather than $18 when it grows, because $24 doubles the RAM for the same CPU money. Budget $0.01 per GiB for anything past the included transfer and treat backups as a 20 to 30 percent surcharge on the whole line (calculated 2026-10-25).

**Sources:**
- DigitalOcean pricing: https://www.digitalocean.com/pricing
- DigitalOcean Droplet pricing: https://www.digitalocean.com/pricing/droplets

---

# 2 vCPU and 4 GB in 2026: $24 on DigitalOcean against $85 on Render

URL: https://actualdatabase.com/digitalocean-vs-render-cost-2vcpu/
Verified: 2026-10-25

**Quick answer:** A 2 vCPU / 4 GB workload costs $24 a month as a DigitalOcean basic Droplet and $85 a month as a Render 2c-4g instance, a gap of $61 a month or $732 a year. The Droplet also includes 4,000 GiB of transfer and 80 GiB of SSD, while Render's Hobby workspace includes 5 GB of bandwidth and charges $0.15 per GB beyond it.

- A 2 vCPU / 4 GB workload costs $24 a month on a DigitalOcean Basic Droplet and $85 a month on Render's 2c-4g instance (checked 2026-10-25).
- That is a $61 monthly gap, or $732 a year, with DigitalOcean about 3.5 times cheaper for identical CPU and RAM (calculated 2026-10-25).
- The Droplet includes 4,000 GiB of outbound transfer and 80 GiB of SSD, while Render's Hobby workspace includes 5 GB of bandwidth (checked 2026-10-25).
- At 100 GB of monthly egress the Render bill becomes about $99.25 once 95 GB of overage at $0.15 per GB is added, against $24 on DigitalOcean (calculated 2026-10-25).
- Render's $85 buys managed TLS, git-based deploys, zero-downtime releases and OS patching, all of which are your work on a bare Droplet (checked 2026-10-25).
- Over the same overage, DigitalOcean charges $0.01 per GiB and Render $0.15 per GiB, a 15 times difference on the marginal gigabyte (checked 2026-10-25).

Four gigabytes of RAM and two vCPUs is the standard middle of a small production stack: an app server with a worker beside it, or one process with enough headroom to survive a deploy. Priced on the two most common platforms, that shape costs $24 a month on a DigitalOcean Basic Droplet and $85 a month on Render's 2c-4g instance. The $61 difference is not a rounding error or a temporary promotion, and it buys two genuinely different products rather than the same machine at two prices.

## The same box, two invoices

| Item | DigitalOcean Droplet | Render 2c-4g |
|---|---|---|
| Monthly price | $24 | $85 |
| vCPU | 2 | 2 |
| RAM | 4 GiB | 4 GB |
| Disk | 80 GiB SSD | ephemeral, disk billed separately |
| Transfer included | 4,000 GiB | 5 GB on Hobby |
| Overage rate | $0.01 per GiB | $0.15 per GB |
| TLS and domains | you configure it | managed, 2 domains on Hobby |
| Deploys | your pipeline | git push, zero downtime |
| OS patching | yours | handled |
| Backups | 20% weekly, 30% daily | database plans include their own |

The RAM units differ by rounding — 4 GiB against 4 GB — and that is the only technical difference between the two allocations. Everything else in the table is operational.

[[ad]]

## What the $61 gap actually buys

Render's premium is priced per gigabyte of RAM, and it scales badly. At 512 MB the platform charges $7 against a $4 to $6 Droplet, a gap of $1 to $3. At 4 GB the same comparison is $85 against $24, a gap of $61. Nothing changed except the size of the machine, which means the platform premium is roughly $15 per gigabyte of RAM at this tier and about $2 per gigabyte at the bottom of the ladder.

Bandwidth widens the gap rather than narrowing it. The Droplet's 4,000 GiB included transfer is more than most apps will ever send, and the marginal gigabyte costs $0.01. Render includes 5 GB on the Hobby workspace and charges $0.15 per GB after it, so a 100 GB month adds $14.25 and takes the total to about $99.25 — more than four times the Droplet (calculated 2026-10-25).

What Render sells for that money is the absence of a certain kind of work: no TLS renewal to wire up, no release script, no patch Tuesday, no SSH keys to rotate. That work is real and it is recurring, but it is also bounded. A team that spends two hours a month on server maintenance is paying $30 an hour for it at the $61 gap, which is a defensible trade; a team that spends twenty minutes is not.

## How we measured it

Three pages were fetched on 2026-10-25: DigitalOcean's Droplet pricing for the $24 tier and its included transfer, DigitalOcean's pricing hub for the $0.01 overage rate and the backup percentages, and Render's pricing page for the 2c-4g instance and the Hobby bandwidth allowance. No workload was deployed on either platform and no invoice was compared, so both totals are list prices and the 100 GB example is arithmetic on published rates.

## What changes in practice

The Droplet price is not the whole Droplet. Backups at 20 or 30 percent take $24 to $28.80 or $31.20, and a managed Postgres instance starts at $15 a month, so the realistic all-in figure for a small production stack is closer to $45 than $24. Render's comparable database plans start at $6 for 256 MB and $19 for 1 GB, which narrows but does not close the gap.

Managed databases are where the two pricing models converge. On DigitalOcean the VM and the database are separate line items that can be sized independently; on Render they are both metered services on the same workspace, which makes the bill simpler and larger. The 2 vCPU machine is cheap on DigitalOcean precisely because it comes with no opinions.

The other difference that bites is disk. A Droplet's 80 GiB SSD is included. On Render the filesystem is ephemeral and a persistent disk costs $0.25 per GB per month, so 80 GiB of durable storage adds $20 — still nowhere near the $61 gap, but it is a line that does not exist on the Droplet invoice.

## Bottom line

Take the $24 Droplet if you can own the operations and want 4,000 GiB of transfer, and take Render's $85 instance only if managed deploys and patching genuinely save more than $61 a month. In between, the arithmetic favours a $24 Droplet plus $15 of managed database at $39, which is still less than half of Render's compute line (calculated 2026-10-25).

**Sources:**
- DigitalOcean pricing: https://www.digitalocean.com/pricing
- DigitalOcean Droplet pricing: https://www.digitalocean.com/pricing/droplets
- Render pricing: https://render.com/pricing

---

# Fastly Pricing: What 1 TB of CDN Traffic Actually Costs

URL: https://actualdatabase.com/fastly-pricing-what-1tb-costs/
Verified: 2026-10-25

**Quick answer:** 1 TB of Fastly traffic from Europe or North America costs $110.88, because the first 100 GB is free and the remaining 924 GB bill at $0.12 per GB. Traffic from Africa costs $258.72 for the same terabyte.

- Fastly charges $0.12 per GB for Full Site Delivery bandwidth in Europe and North America after the first 100 GB free (checked 2026-10-25).
- 1 TB of European or North American bandwidth costs $110.88: 924 GB at $0.12 after the free allowance (calculated 2026-10-25).
- The same terabyte costs $175.56 from Asia, Australia or South America and $258.72 from Africa, India or South Korea (calculated 2026-10-25).
- Fastly publishes no monthly minimum for CDN bandwidth and describes the product as pay-as-you-go; the $500 per month figure on the same page belongs to an edge-compute package (checked 2026-10-25).
- Requests are free to 1 million a month and then $0.01 per 10,000, so 10 million requests add $9.00 (calculated 2026-10-25).
- Bandwidth rates drop past 10 TB per region: $0.08 per GB in Europe and North America, $0.14 in Asia and $0.24 in Africa (checked 2026-10-25).

Fastly prices CDN bandwidth per GB, per region, with the first 100 GB free every month. In Europe and North America the rate is $0.12 per GB, which makes 1 TB cost $110.88 after the free allowance (calculated 2026-10-25). That is roughly ten times what the cheapest CDN charges for the same byte count, so the useful question is not the rate but what the rate includes and which assumptions the page does not state. This page works the arithmetic region by region and records what Fastly does not publish.

## What 1 TB costs by region

| Region | Rate per GB (100 GB–10 TB) | Free allowance | Cost for 1 TB |
|---|---|---|---|
| Europe | $0.12 | 100 GB | $110.88 |
| North America | $0.12 | 100 GB | $110.88 |
| Asia | $0.19 | 100 GB | $175.56 |
| Australia | $0.19 | 100 GB | $175.56 |
| South America | $0.19 | 100 GB | $175.56 |
| Mexico | $0.16 | 100 GB | $147.84 |
| India | $0.28 | 100 GB | $258.72 |
| Africa | $0.28 | 100 GB | $258.72 |
| South Korea | $0.28 | 100 GB | $258.72 |

[[ad]]

## What the published rate actually means

The free allowance is per region per month, not per account, so a site serving 90 GB in Europe and 90 GB in Asia pays nothing for either. The $110.88 figure assumes all 1,024 GB land in one region and the 100 GB credit is consumed once: 924 GB at $0.12.

The rate steps down past 10 TB per region. Europe and North America fall to $0.08 per GB, Asia, Australia, Saudi Arabia and South America to $0.14, Mexico to $0.12 and Africa, India and South Korea to $0.24. Because the tiers are regional, a 15 TB month split evenly across Europe and Asia is billed at the first tier in both regions rather than the second in either.

Requests are a separate meter: 1 million free then $0.01 per 10,000. At 10 million requests the request charge is $9.00, which is noise against bandwidth for video and material for a site serving small icons, where 1 TB can mean tens of millions of requests.

## How we measured it

We loaded the pricing page and the CDN product page on 2026-10-25 and read the Full Site Delivery bandwidth table directly, including the region labels and both volume tiers. The terabyte is read as 1,024 GB and the 100 GB free allowance subtracted once per region. No traffic was served and no account was opened.

One thing we looked for and did not find: a monthly minimum. The pricing page sells "pay-as-you-go, usage-based pricing" and publishes no floor for CDN bandwidth; the only flat monthly figure on it is a $500 per month edge-compute package. Anything quoted as a Fastly minimum elsewhere is not on the page a buyer is linked to.

## What changes in practice

For a small site, Fastly's own free tier is the whole answer: 100 GB of bandwidth and 1 million requests a month cost nothing, which covers a blog with images comfortably and a small SaaS API easily. The first invoice arrives when traffic crosses either line, and it arrives at a rate that assumes enterprise traffic patterns.

The regional table is where budgets break. A product with a global audience can pay two to three times the European rate for the same number of bytes, and the discount for volume only applies inside a region. A 1 TB month in Africa costs $258.72 — more than twice the European price — so a launch that goes viral in the wrong continent can cost more than the servers.

TLS and support are priced separately and are easy to miss. Managed TLS is free for five domains then $20 per domain, a commercial-CA certificate is $275 per domain, and mutual TLS is $500 flat beyond two. Support plans are not on the bandwidth table at all.

## Bottom line

At $110.88 per European or North American terabyte after the 100 GB free allowance, Fastly is worth its price only if the 100 GB free tier covers you or if the platform's other services are already in use. Below 100 GB a month the bill is $0; between 100 GB and 10 TB, budget $0.12 per GB in Europe and North America and up to $0.28 per GB in Africa, India or South Korea.

**Sources:**
- Fastly pricing: https://www.fastly.com/pricing
- Fastly CDN product page: https://www.fastly.com/products/cdn

---

# HubSpot at 1,000 Contacts: Which Tier You Land On and What It Bills

URL: https://actualdatabase.com/hubspot-cost-at-1000-contacts/
Verified: 2026-10-25

**Quick answer:** A 1,000-contact list fits exactly inside the 1,000 marketing contacts included with HubSpot Starter, which costs $20 per seat per month at list, or $7 per seat on the current annual promotion. One seat therefore bills $20 per month, $240 per year. A list of 2,001 contacts leaves Starter behind and lands on Professional at $890 per month with 3 core seats included, plus a one-time $3,000 onboarding fee.

- HubSpot Starter includes 1,000 marketing contacts, exactly covering a 1,000-contact list (checked 2026-10-25).
- Starter costs $20 per seat per month at list, so one seat bills $20 per month or $240 per year for that list (calculated 2026-10-25).
- The current Starter promotion is $7 per seat per month with an annual upfront commitment, which is $84 per year for one seat (calculated 2026-10-25).
- Professional includes 2,000 marketing contacts and costs $890 per month with 3 core seats, plus $3,000 one-time onboarding (checked 2026-10-25).
- Extra marketing contacts are sold in blocks of 5,000, starting at $250 per month for the first block above Professional's included 2,000 (checked 2026-10-25).
- Three seats on Starter for the same 1,000-contact list bill $60 per month at list, because seats multiply and contacts do not (calculated 2026-10-25).

A 1,000-contact list is the exact size where HubSpot's contact allowances start to matter, because it is the last list that fits inside the cheapest paid tier without buying more contacts. Starter includes 1,000 marketing contacts and costs $20 per seat per month, so one seat bills $20 per month, or $240 a year. The current promotion drops the seat to $7 with an annual upfront commitment, which is $84 a year. One contact more than 1,000 and the answer changes completely: Professional starts at $890 per month with 2,000 contacts and 3 core seats, plus a one-time $3,000 onboarding fee.

## Which tier a 1,000-contact list lands on

| Tier | Marketing contacts included | Monthly price, one seat |
| --- | --- | --- |
| Free | Not published on the pricing page | $0 |
| Starter | 1,000 | $20 ($7 promotional) |
| Professional | 2,000 | $890 with 3 core seats |
| Enterprise | 10,000 | $3,600 with 5 core seats |

[[ad]]

## What the same list bills in three configurations

| Configuration | Plan | Monthly bill |
| --- | --- | --- |
| 1,000 contacts, one seat, list price | Starter | $20 |
| 1,000 contacts, one seat, promotional annual | Starter | $7 |
| 1,000 contacts, three seats, list price | Starter | $60 |
| 2,001 contacts, three seats | Professional | $890 plus $3,000 onboarding in year one |

Seat multiplication and contact multiplication are separate (calculated 2026-10-25).

## What the numbers actually mean

Contacts and seats are two independent dials, and only one of them moves with your list. The pricing page sells Starter per seat with 1,000 marketing contacts attached to the plan, so a three-person marketing team with the same 1,000 names pays $60 per month instead of $20 (calculated 2026-10-25) while getting no extra contacts at all. That is the first surprise in a HubSpot bill: headcount, not audience size, drives the ticket until the contact ceiling is crossed.

Crossing 1,000 is the expensive move. Professional is $890 per month, which is 44.5 times the Starter list rate (calculated 2026-10-25), and it arrives with a required $3,000 onboarding fee and 3 core seats. The contact allowance doubles to 2,000 at the same time, so a team at 1,100 contacts pays for a lot of headroom it may never use.

Above Professional's 2,000 included contacts, HubSpot sells blocks of 5,000 rather than individual records. The first block above the allowance is $250 per month, and the per-block price declines through $225, $200 and $175 to $150 once the account passes 82,000 contacts. A 7,000-contact list therefore bills $890 plus $250, or $1,140 per month on the annual rate (calculated 2026-10-25).

There is no separate send fee at any of these levels. HubSpot's Starter documentation states that marketing email has no separate send fee and that the send limit follows the number of marketing contacts purchased, which means a 1,000-contact plan sends to a 1,000-contact audience and the invoice line for email is zero.

## How we measured it

Figures come from the HubSpot Marketing Hub pricing page and the Starter and Professional product pages, all fetched on 2026-10-25, including the structured offer data listing Starter at $20 per seat, Professional at $890 and the 5,000-contact blocks at $250 for the first range. No account was created and no list was uploaded. Seat totals, annual equivalents and multiples are arithmetic on published prices and are labelled calculated.

## What changes in practice

Keep the list under 1,000 while the team is small. That ceiling is worth $10,680 a year in subscription cost at list between Starter and Professional (calculated 2026-10-25), which is more than most small teams spend on the rest of their marketing stack combined.

Buy seats only for people who build campaigns. Because a seat adds $20 per month at list and adds no contacts, a five-person team that only needs two builders is paying $60 per month for read-only colleagues (calculated 2026-10-25), money that would fund two additional 5,000-contact blocks.

Model the year-one bill, not the monthly one. Professional's first year is $12,600 on the annual rate with onboarding (calculated 2026-10-25), and any comparison against a $20 Starter seat should use that number rather than the $890 headline.

## Bottom line

At exactly 1,000 contacts, HubSpot Starter is the whole answer: $20 per month for one seat, $240 a year, or $84 a year on the promotional rate (calculated 2026-10-25). The tier changes at 1,001 contacts, where the bill jumps to $890 per month plus $3,000 onboarding, so managing the list is worth more than any discount you can negotiate.

**Sources:**
- HubSpot Marketing Hub pricing: https://www.hubspot.com/pricing/marketing
- HubSpot Marketing Hub Professional: https://www.hubspot.com/products/marketing/professional
- HubSpot Marketing Hub Starter: https://www.hubspot.com/products/marketing/starter

---

# HubSpot Starter vs Professional: Price, Seats and the Onboarding Fee

URL: https://actualdatabase.com/hubspot-starter-vs-professional/
Verified: 2026-10-25

**Quick answer:** HubSpot Starter is $20 per seat per month at list, currently discounted to $7 per seat with an annual upfront commitment. Professional is $890 per month on Marketing Hub, or $800 billed annually with 3 core seats included, and $100 per seat per month on Sales Hub, or $90 billed annually. Professional also adds a mandatory one-time onboarding fee: $3,000 for Marketing Hub and $1,500 for Sales Hub.

- Marketing Hub Starter is $20 per seat per month at list, discounted to $7 per seat with an annual upfront commitment (checked 2026-10-25).
- Marketing Hub Professional is $890 per month on monthly billing or $800 per month billed annually, and includes 3 core seats (checked 2026-10-25).
- Sales Hub Professional is $100 per seat per month on monthly billing or $90 per seat per month billed annually, against $20 per seat for Starter (checked 2026-10-25).
- Professional adds a required one-time onboarding fee: $3,000 for Marketing Hub and $1,500 for Sales Hub (checked 2026-10-25).
- Three Marketing Hub Professional seats cost $12,600 in year one on annual billing once the $3,000 onboarding fee is included (calculated 2026-10-25).
- Marketing Hub Professional includes 2,000 marketing contacts against 1,000 on Starter, and extra core seats are $45 each per month (checked 2026-10-25).

HubSpot's Starter and Professional tiers are not two sizes of the same product; they are billed on different structures. Starter is $20 per seat per month at list, currently discounted to $7 with an annual upfront commitment. Professional on Marketing Hub is $890 per month, or $800 per month billed annually, and that price already carries 3 core seats. Sales Hub Professional is priced per seat again at $100 per seat per month, or $90 billed annually. On top of both Professional tiers sits a mandatory one-time onboarding fee, $3,000 for Marketing Hub and $1,500 for Sales Hub, which is the line most comparisons leave out.

## Starter and Professional, side by side

| Item | Starter | Professional |
| --- | --- | --- |
| Marketing Hub price | $20 per seat per month ($7 promotional) | $890 per month monthly, $800 billed annually |
| Sales Hub price | $20 per seat per month ($7 promotional) | $100 per seat per month monthly, $90 annually |
| Seats included | One per seat purchased | Marketing: 3 core seats, extra at $45 each |
| Marketing contacts | 1,000 | 2,000 |
| HubSpot credits | 500 per month | 3,000 per month |
| One-time onboarding | None listed | $3,000 Marketing, $1,500 Sales |

[[ad]]

## What three seats cost in year one

| Cost line | Marketing Hub | Sales Hub |
| --- | --- | --- |
| Starter, 3 seats at list | $720 per year | $720 per year |
| Professional, 3 seats, annual billing | $9,600 per year | $3,240 per year |
| One-time onboarding | $3,000 | $1,500 |
| Year-one total, annual billing | $12,600 | $4,740 |

All totals are arithmetic on the published rates (calculated 2026-10-25).

## What the numbers actually mean

The Marketing Hub jump is the expensive one. Three seats on Starter cost $720 a year at list, while Professional costs $12,600 in year one, a multiple of 17.5 (calculated 2026-10-25). Of that total, $3,000 is onboarding paid once and $9,600 is subscription, so the recurring bill after year one is $9,600 rather than $12,600 on annual billing.

Money that Marketing Professional charges per platform rather than per seat changes the shape of the bill. The $890 covers three core seats, so a three-person team pays $296.67 per seat per month, while a six-person team on the same plan pays $1,025 per month, or $170.83 per seat (calculated 2026-10-25). Starter moves in the opposite direction: every seat is a full $20, so a ten-person team on Starter pays $200 per month while a ten-person team on Professional pays $890 plus $315 for the seven extra core seats, or $1,205 (calculated 2026-10-25).

Sales Hub is the cheaper path into Professional. At $90 per seat on annual billing, three seats are $3,240 a year against $9,600 for Marketing, and the onboarding fee is half the size. A team that mainly needs pipeline stages, sequences and forecasting rather than email campaigns can reach Professional without the Marketing Hub price.

The discount is a timing trap. Starter's $7 figure requires upfront payment on an annual commitment, and the same product is $10 per seat per month without a commitment against a $20 list price. Any plan that assumes Starter at $7 should also assume it can end.

## How we measured it

Prices come from the HubSpot Marketing Hub and Sales Hub pricing pages fetched on 2026-10-25, plus the Marketing Hub Starter product page for the promotional terms. Figures were cross-checked against the structured offer data embedded in those pages, which lists Marketing Professional at $890, Sales Professional at $100 and Enterprise at $3,600 and $150. No trial was started and no quote was requested. Seat totals and multi-year sums are arithmetic on published rates and are marked calculated.

## What changes in practice

Price the onboarding fee into year one or the comparison is wrong. Professional on Marketing Hub is $9,600 a year in subscription, so a $3,000 one-time fee is a 31% surcharge on the first year (calculated 2026-10-25) and nothing in year two.

Do not compare per-seat numbers across tiers. Marketing Professional's per-seat cost depends on team size because three seats are bundled, which means a five-person team really pays $980 per month, or $196 per seat, not $890 (calculated 2026-10-25). Starter is the only tier where per-seat pricing means what it says.

Treat the $7 promotion as a trial price. Budget the $20 list rate for anything longer than the promotional period, and keep the annual commitment question separate from the tier decision, because mixing the two is how teams end up locked into Professional at list.

## Bottom line

For three seats, Starter is $720 a year and Marketing Professional is $12,600 in year one, or $9,600 a year after that (calculated 2026-10-25). Buy Professional when you need its automation and reporting and can absorb the $3,000 onboarding, and check Sales Hub first: at $4,740 for three seats in year one it is under half the Marketing Hub equivalent.

**Sources:**
- HubSpot Marketing Hub pricing: https://www.hubspot.com/pricing/marketing
- HubSpot Sales Hub pricing: https://www.hubspot.com/pricing/sales
- HubSpot Marketing Hub Starter: https://www.hubspot.com/products/marketing/starter

---

# JetBrains All Products Pack: Price, Per-Product Alternative, Year-Two Curve

URL: https://actualdatabase.com/jetbrains-all-products-pack-cost/
Verified: 2026-10-25

**Quick answer:** The All Products Pack is 2,500 CNY per year in year one, 2,000 CNY in year two and 1,500 CNY from year three, which is the standard 0/20/40 continuing-discount curve. A single IntelliJ IDEA Ultimate subscription is 1,650 CNY in year one and falls to 1,320 and then 990. Two products still cost less than the pack in every year, so the pack pays off at three products, not two.

- The All Products Pack is 2,500 CNY for the first year of a personal annual subscription (checked 2026-10-25).
- The same pack drops to 2,000 CNY in year two and 1,500 CNY from year three, a 20% then 40% continuing discount (checked 2026-10-25).
- IntelliJ IDEA Ultimate on its own is 1,650 CNY in year one, 1,320 in year two and 990 from year three (checked 2026-10-25).
- The pack costs 1.52 times a single IDEA Ultimate subscription in year one, and by year three the 1,500 CNY pack is cheaper than a first-year 1,650 CNY IDE (calculated 2026-10-25).
- Two personal annual subscriptions, IDEA Ultimate at 1,650 plus WebStorm at 650, total 2,300 CNY against 2,500 CNY for the pack in year one (calculated 2026-10-25).
- Commercial All Products Pack is 8,080 CNY per year or 808 CNY per month, against 5,930 CNY a year for commercial IDEA Ultimate alone (checked 2026-10-25).

JetBrains sells one subscription per product and one subscription for everything, and the price difference between them is smaller than most developers assume. The All Products Pack is 2,500 CNY for the first year of a personal annual subscription, then 2,000 CNY in year two and 1,500 CNY from year three. A single IntelliJ IDEA Ultimate subscription is 1,650 CNY, 1,320 and 990 over the same three years. The pack costs 1.52 times one IDE in year one (calculated 2026-10-25), which sounds like a lot until you price two products, and the year-three figures reverse the intuition entirely.

One note before the numbers: the JetBrains store quotes prices for the country it geolocates. The store data captured for this page was set to China Mainland, so every figure below is in CNY, at the personal or commercial rate the page lists. The percentage curve is currency-independent; the absolute numbers are not.

## Personal annual subscriptions, three years side by side

| Subscription (personal, yearly) | Year 1 | Year 2 | Year 3 onward |
| --- | --- | --- | --- |
| All Products Pack | 2,500 CNY | 2,000 CNY | 1,500 CNY |
| IntelliJ IDEA Ultimate | 1,650 CNY | 1,320 CNY | 990 CNY |
| PhpStorm | 890 CNY | 712 CNY | 534 CNY |
| WebStorm | 650 CNY | 520 CNY | 390 CNY |

[[ad]]

## When the pack beats separate subscriptions

| Basket of separate subscriptions | Year 1 total | Against the 2,500 CNY pack |
| --- | --- | --- |
| IDEA Ultimate alone | 1,650 CNY | Pack costs 850 CNY more |
| IDEA Ultimate plus WebStorm | 2,300 CNY | Pack costs 200 CNY more |
| IDEA Ultimate plus WebStorm plus PhpStorm | 3,190 CNY | Separate costs 690 CNY more |

Totals are sums of the published first-year prices (calculated 2026-10-25).

## What the numbers actually mean

The pack only wins at three products. Two subscriptions, IDEA Ultimate at 1,650 plus WebStorm at 650, come to 2,300 CNY against 2,500 CNY, and the gap does not close with tenure: year two is 1,840 against 2,000 and year three is 1,380 against 1,500 (calculated 2026-10-25). The reason is that the pack applies the same 0/20/40 curve to a larger base, so its discount saves more currency but never changes the ordering.

Add one more IDE and the ordering flips. IDEA Ultimate, WebStorm and PhpStorm total 3,190 CNY in year one against 2,500 for the pack, and the pack keeps winning in later years (calculated 2026-10-25). Anyone who uses three JetBrains tools, even occasionally, is paying a premium for buying them separately, which is exactly what the pack is priced to capture.

The curve itself is the more interesting figure. Personal annual prices step 0%, 20%, 40% across the three years, so a pack subscription costs 60% of list by year three: 1,500 CNY against 2,500 (calculated 2026-10-25). Cancelling and restarting resets the clock to full price, which makes the discount a loyalty asset rather than a discount anyone can re-earn.

Commercial rates run on a different ladder, without the visible multi-year step on the store page: the All Products Pack is 8,080 CNY per year, or 673.33 CNY per month averaged, and 808 CNY on monthly billing, while commercial IDEA Ultimate alone is 5,930 CNY per year (calculated 2026-10-25). The pack is 1.36 times a single commercial IDE, a narrower multiple than the personal ratio.

## How we measured it

Figures come from the JetBrains store page fetched on 2026-10-25, whose embedded product data lists, for each subscription, the yearly price for year one, two and three with the matching percentage off, plus the monthly billing rate. No license was purchased, no trial was run and no tool was benchmarked. Basket totals, ratios and per-month averages are arithmetic on those published prices and are marked calculated.

## What changes in practice

Count the tools before buying. Two JetBrains products, even two expensive ones, are cheaper bought separately in every year of the curve; the pack becomes the rational purchase at the third product or when the team expects to add one.

Do not cancel to re-shop. The 40% year-three rate on the personal pack is worth 1,000 CNY a year against list, and a lapsed subscription restarts at 2,500 CNY (calculated 2026-10-25). The cheapest path is continuous renewal, not an annual comparison shop.

Budget the third year, not the first. A three-year pack commitment costs 6,000 CNY in total at the published rates, an average of 2,000 CNY a year (calculated 2026-10-25), so a first-year budget of 2,500 overstates the steady-state cost and understates the lock-in.

## Bottom line

At 2,500 CNY for year one and 1,500 CNY from year three, the All Products Pack is worth it once you would otherwise buy three subscriptions; at two, separate licenses save 200 CNY in year one and 120 CNY in year three (calculated 2026-10-25). Everyone else should buy the one IDE they actually use and let the discount curve run.

**Sources:**
- JetBrains store pricing: https://www.jetbrains.com/store/
- JetBrains licensing FAQ: https://sales.jetbrains.com/hc/en-gb

---

# Lemon Squeezy Fees Per Sale in 2026: 5% + 50¢ and What It Covers

URL: https://actualdatabase.com/lemonsqueezy-fees-per-sale-2026/
Verified: 2026-10-25

**Quick answer:** Lemon Squeezy charges 5% + 50¢ per transaction with no monthly fee, taken from the total order value. A $10 sale costs $1.00 and nets $9.00; a $100 sale costs $5.50 and nets $94.50.

- Lemon Squeezy charges 5% + 50¢ per transaction on the total order value, with no monthly fee (checked 2026-10-25).
- A $10 sale incurs $1.00 and nets $9.00; a $100 sale incurs $5.50 and nets $94.50 (calculated 2026-10-25).
- The platform fee includes card processing, currency conversion and tax handling because Lemon Squeezy is the merchant of record (checked 2026-10-25).
- A $10 product sold to a French buyer at 20% VAT totals $12.00, incurs $1.28 and nets $8.72 after tax (calculated 2026-10-25).
- Payouts are free to US bank accounts through Stripe and cost 1% outside the US; PayPal payouts are $0.50 flat in the US and 3% capped at $30 outside it (checked 2026-10-25).
- International buyers add 1.5%, PayPal adds 1.5% and subscription payments add 0.5% on top of the base fee (checked 2026-10-25).

Lemon Squeezy charges 5% + 50¢ per transaction on the total order value, on a $0/month plan. A $10 sale therefore costs $1.00 and nets $9.00; a $100 sale costs $5.50 and nets $94.50 (calculated 2026-10-25). That is 70% more than Stripe's 5.9% on the same $10 charge, and the reason is that Lemon Squeezy is the merchant of record, not just the payment form. This page takes the fee schedule as published on 2026-10-25 and separates what the fee buys from what it costs.

## What a sale costs on Lemon Squeezy

| Order total | Fee at 5% + 50¢ | You keep | Fee as share of sale |
|---|---|---|---|
| $10 | $1.00 | $9.00 | 10.0% |
| $20 | $1.50 | $18.50 | 7.5% |
| $50 | $3.00 | $47.00 | 6.0% |
| $100 | $5.50 | $94.50 | 5.5% |
| $500 | $25.50 | $474.50 | 5.1% |

[[ad]]

## What the fee actually includes

The published fee is a single line, but it replaces four separate purchases: card processing, currency conversion, tax calculation and tax remittance. Lemon Squeezy states plainly that the platform fee covers taxes, and that as merchant of record it carries the liability for collecting, calculating and filing them. A merchant who sells a $10 download to a buyer in France does not file a French VAT return, because Lemon Squeezy is the seller on the invoice.

| Item | Rate | When it applies |
|---|---|---|
| Plan fee | $0/month | Always |
| Transaction fee | 5% + 50¢ | Every sale, on the total |
| International | +1.5% | Buyer outside the US |
| PayPal | +1.5% | PayPal as the payment method |
| Subscriptions | +0.5% | Recurring charges |
| Payout, US Stripe | $0.00 | US bank account |
| Payout, non-US Stripe | 1% | Per payout |
| Payout, PayPal | $0.50 US, 3% capped $30 elsewhere | Per payout |

The tax line is where the arithmetic gets interesting. A $10 product sold into France carries 20% VAT, so the buyer pays $12.00 and $2.00 of that is tax. The fee is calculated on the $12.00 total, not on the $10.00 net: $0.50 fixed, 5% ($0.60), and 1.5% international ($0.18) come to $1.28, leaving $8.72 of your $10.00 (calculated 2026-10-25). Tax raises the fee because the fee is charged on the gross.

## How we measured it

We loaded the pricing page and the fees help article on 2026-10-25 and used only rates printed there, including the vendor's own worked example for a French buyer. Where the page states a rule in prose — custom pricing below $10, tax liability sitting with them — we repeated the rule rather than inventing a number. No benchmark was run and no test checkout was created.

## What changes in practice

At $10 the fee is 10% of the sale, which is the worst ratio on the table. It falls to 7.5% at $20 and 5.5% at $100, so the same software priced at $20 instead of $10 keeps 83% more per sale in absolute terms. Below $10 the company itself points you at custom pricing, which is a polite way of saying the standard fee is not built for that ticket size.

The surcharges are easy to forget because they are not in the headline. A subscription sold to a European buyer paying with PayPal costs 5% + 50¢ + 1.5% + 1.5% + 0.5%, which on $10 is $1.35 and leaves $8.65 — against $9.00 for the plain US card case.

Payouts are the second cost centre, and they scale with frequency rather than volume. A merchant outside the US paying out weekly pays 1% each time; someone using PayPal payouts abroad pays 3% capped at $30, so a $1,000 payout costs $30 while a $200 payout costs $6.00.

## Bottom line

At a $10 price point Lemon Squeezy keeps $1.00 of every sale, and at $20 it keeps $1.50 — the fee only becomes competitive past roughly $100, where 5.5% is within two points of what a self-managed tax stack costs in time. If you sell digital goods to consumers in several tax jurisdictions and do not want to register for VAT yourself, that premium buys a real service; if you sell to businesses in one country, it buys very little.

**Sources:**
- Lemon Squeezy pricing: https://www.lemonsqueezy.com/pricing
- Lemon Squeezy docs: fees: https://docs.lemonsqueezy.com/help/getting-started/fees

---

# Mailchimp Free Plan Limits in 2026: 250 Contacts, 500 Sends, 30 Days of Support

URL: https://actualdatabase.com/mailchimp-free-plan-limits-2026/
Verified: 2026-10-25

**Quick answer:** The Mailchimp free plan includes 1 seat, 1 audience with a maximum of 250 contacts, and up to 500 sends per month with a daily cap of 250. Support is email only, for the first 30 days. The first paid step is Essentials, free for 14 days and then starting at $13 per month.

- Mailchimp Free allows 1 seat and 1 audience with a maximum of 250 contacts (checked 2026-10-25).
- The free plan allows up to 500 sends per month, with a daily limit of 250 sends (checked 2026-10-25).
- Free plan support is email only and lasts for the first 30 days (checked 2026-10-25).
- The first paid step is Essentials, free for 14 days and then starting at $13 per month (checked 2026-10-25).
- Paid plans scale the send limit with contacts: 10 times the contact count on Essentials, 12 times on Standard and 15 times on Premium (checked 2026-10-25).
- At 250 free contacts, two monthly campaigns use the entire 500-send allowance, so four campaigns overrun the monthly cap by 500 sends (calculated 2026-10-25).

Mailchimp's free plan is capped in four places at once, and the cap that stops most senders is not the one they expect. The plan includes 1 seat, 1 audience with a maximum of 250 contacts, and up to 500 sends per month, with a daily ceiling of 250 sends. Support is email only and lasts the first 30 days. Once any of those runs out, the first paid step is Essentials, which is free for 14 days and then starts at $13 per month.

## The four free plan limits

| Limit | Free plan | Where the next tier differs |
| --- | --- | --- |
| Contacts | 250 maximum, 1 audience | Essentials and above scale the allowance, listed up to 50,000 contacts |
| Monthly sends | 500, no rollover stated | 10 times the contact count on Essentials, 12 times on Standard, 15 times on Premium |
| Daily sends | 250 | Not capped by the same rule on paid plans |
| Seats | 1 | Additional seats on paid tiers |
| Support | Email, first 30 days | Email and chat on Essentials and Standard, phone on Premium |
| Price | $0 | Essentials from $13 per month after a 14-day trial |

[[ad]]

## What 500 sends a month allows at 250 contacts

| Campaigns per month to the full list | Sends used | Against the 500 monthly cap |
| --- | --- | --- |
| 1 | 250 | 250 left |
| 2 | 500 | Exactly at the cap |
| 4 | 1,000 | 500 over |
| 8 | 2,000 | 1,500 over |

Send counts assume every campaign goes to all 250 contacts (calculated 2026-10-25).

## What the numbers actually mean

The send cap binds before the contact cap. At 250 contacts, two monthly campaigns to the whole list consume the entire 500-send allowance, which means a sender who mails weekly is two or three campaigns short every month (calculated 2026-10-25). The contact limit only becomes the binding constraint for a list that is mailed once or twice a month, and almost no newsletter is.

That is why the free plan feels tighter than 250 contacts suggests. A 250-contact list mailed weekly sends roughly 1,000 messages a month and needs twice the allowance, so at four campaigns a month the real ceiling is a 125-contact list (calculated 2026-10-25). Anyone planning a weekly send should size the list against sends, not contacts.

The support window is the limit nobody prices. Email support ends after 30 days on Free, and there is no phone support anywhere below Premium, so a new sender gets one month of help and then self-service. For a paid plan that matters less, because Essentials and Standard both include email and chat support for as long as the subscription runs.

The gaps between paid tiers are set by a multiplier rather than a fixed number. Essentials allows 10 times the contact count in monthly sends, Standard 12 times and Premium 15 times. At 1,000 contacts that is 10,000 sends on Essentials and 15,000 on Premium (calculated 2026-10-25), which is the practical reason to move up a tier before the contact count forces it.

## How we measured it

Every limit here comes from the Mailchimp pricing page fetched on 2026-10-25, where the free plan terms appear in the plan tooltips and the send multipliers appear in the plan comparison copy. The page is a JavaScript calculator and does not print a per-contact price table in its HTML, so cost figures are limited to the published starting prices: $13 per month for Essentials, $20 per month for Standard, and custom pricing for Premium. Multipliers, send totals and list sizes are arithmetic on those published rules and are marked calculated.

## What changes in practice

Size the list to the cadence you actually send. A weekly newsletter on the free plan caps out around 125 contacts, and a monthly one reaches the full 250, so the plan's real usefulness depends on how often you hit send.

Watch the daily cap when importing. 250 sends per day means a 250-contact list can be mailed once per day, not twice, and a list of 200 contacts can take a second campaign only by splitting the day.

Plan the upgrade before the cap bites. Essentials at $13 per month buys 10 times the contact count in sends, which at 250 contacts is 2,500 sends, five times the free allowance (calculated 2026-10-25). Moving up is cheaper than running two free accounts, which the single-audience rule does not allow anyway.

## Bottom line

The free plan is real but narrow: 250 contacts, 500 sends a month, one seat and 30 days of email support. If you send weekly, plan on 125 contacts or on the $13 Essentials tier, because at 250 contacts a weekly cadence needs 1,000 sends against a 500-send cap (calculated 2026-10-25).

**Sources:**
- Mailchimp pricing: https://mailchimp.com/pricing/
- Mailchimp plan comparison: https://mailchimp.com/pricing/marketing/compare-plans/

---

# Mailchimp vs Buttondown at 1,000 Subscribers and Four Sends a Month

URL: https://actualdatabase.com/mailchimp-vs-buttondown-for-1000-subscribers/
Verified: 2026-10-25

**Quick answer:** Buttondown publishes the number: $9 per month for up to 1,000 subscribers, which is $108 a year, and four sends a month sits inside its one-email-a-day assumption. Mailchimp does not publish a 1,000-contact price in its served pages; the published floor for a list that size is Essentials at $13 per month after a 14-day trial, with Standard from $20 per month, and the exact 1,000-contact figure comes only from Mailchimp's on-page calculator.

- Buttondown Basic costs $9 per month for up to 1,000 subscribers (checked 2026-10-25).
- That is $108 per year for 1,000 subscribers, or $54 a year with Buttondown's 50% non-profit discount (calculated 2026-10-25).
- Buttondown's free tier stops at 100 subscribers and Standard, at $29 per month, covers up to 5,000 (checked 2026-10-25).
- Mailchimp's free plan stops at 250 contacts and 500 sends per month, so 1,000 subscribers is a paid tier either way (checked 2026-10-25).
- Mailchimp's published paid floor is Essentials at $13 per month after a 14-day trial, with Standard starting at $20 per month (checked 2026-10-25).
- Four sends to 1,000 subscribers is 4,000 emails a month, inside Mailchimp Essentials' 10-times-contacts send allowance of 10,000 (calculated 2026-10-25).

At 1,000 subscribers and four sends a month, one of these two vendors tells you the price and the other makes you calculate it. Buttondown publishes $9 per month for up to 1,000 subscribers, which is $108 a year, and its prices assume at most one email a day, so four monthly sends fit. Mailchimp publishes no 1,000-contact figure in the pages it serves: the free plan ends at 250 contacts and 500 sends, Essentials starts at $13 per month after a trial, and the exact number for a 1,000-contact list only appears once you move the calculator.

## What each vendor publishes for 1,000 subscribers

| Item | Buttondown | Mailchimp |
| --- | --- | --- |
| Free tier ceiling | 100 subscribers | 250 contacts, 500 sends per month |
| Price for 1,000 subscribers | $9 per month, $108 per year | Not published; Essentials from $13 per month, Standard from $20 |
| Next tier up | Standard, $29 per month, up to 5,000 | Contact ladder, up to 50,000 contacts at $385 per month |
| Send limit basis | Price assumes at most one email a day | 10 times contacts on Essentials, 12 times on Standard |
| Discounts listed | 50% for non-profits | 15% off the first 12 months at 10,000 or more contacts |

[[ad]]

## What four sends a month does to each plan

| Question | Buttondown at $9 | Mailchimp Essentials |
| --- | --- | --- |
| Emails sent per month at 1,000 subscribers | 4,000 | 4,000 |
| Monthly send allowance | Not metered separately; one email a day assumed | 10,000, which is 10 times 1,000 contacts |
| Does the cadence change the price? | No, inside the stated assumption | No, 4,000 sends sits under the allowance |
| Annual cost at four sends a month | $108 | Above $156, exact figure from the calculator |

The send count is 1,000 subscribers times four campaigns (calculated 2026-10-25); the Mailchimp annual figure is $13 times twelve as a floor, not a quote.

## What the numbers actually mean

Buttondown's price is per subscriber ceiling, not per send. Paying $9 for up to 1,000 subscribers means a 400-subscriber list costs the same as a 1,000-subscriber one, and sending four times a month is inside the stated assumption of at most one email a day. The bill only moves at 100, 1,000, 5,000, 10,000 and 20,000 subscribers, which makes it predictable for a newsletter that grows steadily.

Mailchimp's structure is the opposite: the price tracks contacts, and the send limit is a multiple of contacts rather than a fixed number. On Essentials that multiple is 10, so 1,000 contacts buys 10,000 monthly sends, and four campaigns of 1,000 use 4,000 of them (calculated 2026-10-25). The cadence is not what costs money on Mailchimp; the list size is.

The unpriced gap is the story. Mailchimp's free plan covers 250 contacts and 500 sends, and its next published rung is 50,000 contacts at $385 per month, so the 250-to-50,000 range is quoted only by the calculator. Anyone budgeting $13 for a 1,000-contact list is using the entry price of the smallest paid tier as if it were a quote for their tier, which is the most common arithmetic error in these comparisons.

Engagement rules diverge too. Buttondown states it charges for active subscribers, so a list with heavy churn or a large inactive tail bills on the live portion, while Mailchimp counts contacts in the audience against the plan. On a 1,000-name list with 20% dead addresses, Buttondown bills 800 subscribers inside the same $9 tier while Mailchimp counts all 1,000 toward its allowance.

## How we measured it

Figures come from three pages fetched on 2026-10-25: Mailchimp's pricing page, Mailchimp's plan comparison page, and Buttondown's pricing page, whose structured product data lists Free at $0 for 100 subscribers, Basic at $9 for 1,000, Standard at $29 for 5,000, Professional at $79 for 10,000 and Advanced at $139 for 20,000. No account was opened on either service and no campaign was sent. Annual totals and send counts are arithmetic on published prices and are labelled calculated.

## What changes in practice

If the list is between 100 and 1,000 subscribers and the cadence is weekly, Buttondown is the only one of the two with a published price for that size, and it is $9 per month. That is a budget line, not an estimate.

If the list is heading past 1,000, run Mailchimp's calculator before comparing anything, because the $13 figure will not survive contact with a 5,000-contact list. The published rung at 50,000 contacts is $385 per month, so the ladder rises steeply between the two.

If support matters, read the support rows rather than the price. Mailchimp's free plan stops at 30 days of email support and its paid tiers include email and chat, while Buttondown's pricing page lists no support tier at all, so the $9 buys the product and nothing documentable beyond it.

## Bottom line

At 1,000 subscribers and four sends a month, Buttondown's published bill is $9 per month, $108 a year (calculated 2026-10-25). Mailchimp's published floor is $13 per month and its real 1,000-contact price is not on the page, so the honest comparison is $108 against something above $156 until the calculator produces the number.

**Sources:**
- Mailchimp pricing: https://mailchimp.com/pricing/
- Mailchimp plan comparison: https://mailchimp.com/pricing/marketing/compare-plans/
- Buttondown pricing: https://buttondown.com/pricing

---

# Mailgun free plan limits: 100 emails a day and one domain

URL: https://actualdatabase.com/mailgun-free-plan-limits/
Verified: 2026-10-25

**Quick answer:** Mailgun's free plan allows 100 emails a day — roughly 3,000 a month — on one custom sending domain, with 2 API keys, 1 inbound route, 1 day of log retention and 1 user per account. The pricing page does not describe what happens when the 100-a-day cap is reached; what it does state is that Basic at $15/month removes the daily limit and includes 10,000 emails.

- Mailgun's free plan costs $0/month and includes 100 emails per day (checked 2026-10-25).
- The free plan includes 1 custom sending domain, 2 API keys, 1 inbound route, 1 day of log retention and 1 user per account (checked 2026-10-25).
- 100 emails a day works out to about 3,000 emails in a 30-day month, or 3,100 in a 31-day month (calculated 2026-10-25).
- Mailgun Basic costs $15/month, includes 10,000 emails, states no daily email limit and charges $1.80 per 1,000 extra (checked 2026-10-25).
- The free card is labelled 'Test us out for $0/mo', and Mailgun's FAQ describes the free plan as a way to explore the platform with a daily sending cap (checked 2026-10-25).
- The pricing page does not state what happens to messages once the free plan's 100-a-day cap is reached (checked 2026-10-25).

Mailgun's free plan is one of the few that still gives a real email API away, and its shape is unusual: a daily ceiling rather than a monthly bucket. That makes it fine for a side project and awkward for anything with a weekly batch.

The plan includes 100 emails a day — about 3,000 a month — on one custom sending domain, with 2 API keys, 1 inbound route, 1 day of logs and 1 user per account. The pricing page does not say what happens at the cap; it says Basic at $15 removes the daily limit.

## What the free plan includes

| Item | Free ($0) | Basic ($15) | Foundation ($35) |
|---|---|---|---|
| Included email | 100 per day | 10,000 per month | 50,000 per month |
| Daily limit | 100 | None | None |
| Custom domains | 1 | 1 | 1,000 |
| Log retention | 1 day | 1 day | 5 days |
| Inbound routes | 1 | 5 | Not listed |
| Extra per 1,000 | Not offered | $1.80 | $1.30 at 50k |

All three tiers include the REST API and SMTP relay, tracking, analytics and webhooks, and 2 API keys. Support on the free plan is ticket-based.

[[ad]]

## What the numbers actually mean

A daily cap and a monthly cap are not the same product. 100 a day is 3,000 a month if you send every day, but it is 100 a month if you only send on launch day, and no unused allowance carries forward. A password-reset flow that peaks at 400 messages on a Monday cannot be served by the free plan at any monthly volume.

The feature list is where the real ceiling sits. One custom sending domain means one product, one brand and one set of DNS records; a second domain requires upgrading. One day of log retention means a delivery problem investigated two days later has no evidence behind it, which is the constraint most teams notice after an incident rather than before.

The jump to Basic is small in money and large in shape. $15 buys 10,000 emails a month with no daily limit, five inbound routes, and an overage rate of $1.80 per 1,000 — 3.3 times the free plan's monthly ceiling, at a price that stays flat whether you send 500 emails or 10,000.

A daily cap changes what a launch looks like. Sending 2,500 emails in one afternoon would need 25 days of free-plan allowance, so the constraint is the peak rather than the monthly average — the same argument as a bandwidth cap, and the reason bursty transactional traffic is the wrong workload for a plan with a daily ceiling.

## How we measured it

Plan inclusions, limits and prices are quoted from Mailgun's pricing page on 2026-10-25; the framing of the free plan as a way to explore the platform comes from the FAQ on that page. The monthly equivalents of the daily cap are arithmetic on the published 100-a-day figure and are marked calculated. Nothing was sent, so the behaviour at the cap is reported as undocumented rather than guessed at.

## What changes in practice

The first wall is the daily cap, and it arrives without a monthly dashboard warning. Transactional traffic is bursty by nature — a deploy that re-sends verification emails, a retry storm, a batch import — and a 100-a-day ceiling is reached in minutes at any real scale.

The second wall is the single domain. A product with a marketing domain and a transactional subdomain, or a team running two brands, needs two custom sending domains and therefore a paid plan regardless of volume.

The third is support and observability. One day of retention on a free account means you cannot answer "did that email bounce on Friday" on Monday, and ticket support on a $0 plan is not an incident channel.

Trying the next tier is cheap. Mailgun's FAQ states that paid plans include free 30-day trials with access to higher tier features, so a team can run Foundation at $35 for a month and compare its five-day retention and five inbound routes against the free plan before committing.

## Bottom line

Mailgun's free plan is worth using exactly while you send under 100 emails a day, run one domain, and can live with one day of logs. The moment any of those three stops being true, Basic at $15 for 10,000 emails a month is the cheaper answer than engineering around the cap — and it removes the daily limit entirely, which is the limit that actually breaks things.

**Sources:**
- Mailgun pricing: https://www.mailgun.com/pricing/
- Mailgun sending platform: https://www.mailgun.com/products/send/

---

# Mailgun vs Postmark: what 20,000 emails a month actually costs

URL: https://actualdatabase.com/mailgun-vs-postmark-pricing/
Verified: 2026-10-25

**Quick answer:** 20,000 emails a month costs $33.00 on Mailgun Basic and $29.50 on Postmark Pro, which is the cheapest option either vendor publishes at that volume. Postmark's entry Basic plan also lands at $33.00, and Mailgun's $35 Foundation tier only beats its own Basic tier above about 21,100 emails a month.

- Mailgun Basic costs $15/month for 10,000 emails and charges $1.80 per 1,000 extra (checked 2026-10-25).
- Postmark Basic costs $15/month for 10,000 emails at $1.80 per 1,000 extra; Pro is $16.50 at $1.30 and Platform is $18.00 at $1.20 (checked 2026-10-25).
- 20,000 emails a month costs $33.00 on Mailgun Basic, $33.00 on Postmark Basic, $29.50 on Postmark Pro and $30.00 on Postmark Platform (calculated 2026-10-25).
- Mailgun Foundation includes 50,000 emails for $35/month, which overtakes Mailgun Basic at about 21,111 emails a month (calculated 2026-10-25).
- Mailgun's free plan allows 100 emails per day; Postmark's free developer plan allows 100 emails per month with no overages allowed (checked 2026-10-25).
- Postmark states it offers no tier between 100 emails a month and 10,000 emails a month (checked 2026-10-25).

Both vendors quote $15 for 10,000 emails, which makes them look interchangeable at the entry point. The difference appears at the overage line, where one vendor sells a second tier at a lower marginal rate and the other sells a bigger bucket.

At 20,000 emails a month the totals are $33.00 on Mailgun Basic, $33.00 on Postmark Basic, $29.50 on Postmark Pro and $30.00 on Postmark Platform. Mailgun's $35 Foundation tier includes 50,000 emails and only wins above roughly 21,100.

## Entry tiers and what 20,000 emails costs

| Plan | Monthly fee | Included | Extra per 1,000 | 20,000 emails |
|---|---|---|---|---|
| Mailgun Basic | $15.00 | 10,000 | $1.80 | $33.00 |
| Mailgun Foundation | $35.00 | 50,000 | $1.30 at 50k | $35.00 |
| Postmark Basic | $15.00 | 10,000 | $1.80 | $33.00 |
| Postmark Pro | $16.50 | 10,000 | $1.30 | $29.50 |
| Postmark Platform | $18.00 | 10,000 | $1.20 | $30.00 |

Mailgun also sells Scale at $90 for 100,000 emails with a sliding overage rate that falls to $0.50 per 1,000 at a million. Postmark has no equivalent volume tier, which is why its three plans all sit between $15 and $18.

[[ad]]

## What the numbers actually mean

The two vendors price the same way at the bottom and diverge immediately above it. Mailgun's answer to needing more than 10,000 emails is a larger included bucket: Foundation at $35 for 50,000. Postmark's answer is a better marginal rate on the same bucket: Pro at $16.50 with $1.30 per 1,000 extra.

That makes Postmark cheaper for the band between 10,000 and 50,000 emails a month, and Mailgun cheaper from the point where Foundation's flat $35 undercuts a metered bill. At 40,000 emails, Postmark Pro is $16.50 + 30 x $1.30 = $55.50 while Mailgun Foundation is $35.00. At 20,000 the ordering reverses: $29.50 against $35.00.

The per-1,000 view shows how each vendor moves. Mailgun Basic and Postmark Basic both charge $1.50 per 1,000 at their 10,000-email allowance and $1.80 beyond it. Postmark then cuts the marginal rate instead of the base — $1.30 on Pro, $1.20 on Platform — which is why Mailgun's Foundation tier wins at 50,000 emails with a flat $35.00 while Postmark pays $66.00, and why Postmark reaches $126.00 at 100,000 emails where Mailgun's Scale tier is $90.00.

The free tiers are not comparable products. Mailgun gives 100 emails a day, which is a development allowance that can carry a small production site. Postmark gives 100 emails a month and states plainly that no overages are allowed and that it sells nothing between that and 10,000 — so the real entry price at Postmark is $15, not $0.

## How we measured it

Every price, allowance and overage rate is quoted from the Mailgun and Postmark pricing pages, both opened on 2026-10-25. Postmark's page prices plans against a volume selector that defaults to 10,000 emails, so the 20,000-email column was computed from the published per-1,000 rate rather than read off the page; those figures are marked calculated. Nothing was sent and no deliverability test was run.

## What changes in practice

Per-1,000 pricing hides how billing actually behaves. Postmark charges overage per 1,000 emails and does not roll unused emails over between months, so a seasonal sender pays for the peak month and starts the quiet month from zero.

Mailgun's tier structure is worth checking before a launch, because the next bucket up can be cheaper than the overage on the current one. Basic at 25,000 emails is $15 + 15 x $1.80 = $42.00, while Foundation is $35.00 for 50,000 — the upgrade pays for itself and adds headroom.

Both vendors gate the features that matter at volume. Postmark offers dedicated IPs only from 300,000 emails a month, and Mailgun's dedicated IP pools sit on the Scale tier at $90 a month, so the headline per-1,000 rates do not describe what a serious sender ends up paying.

Retention separates them as capability rather than money. Postmark keeps 45 days of message data on Basic and lets Pro extend it to a customizable 365 days, while Mailgun keeps one day on Free and Basic, five days on Foundation and thirty on Scale. That choice decides whether last month's delivery problem can still be investigated at all.

## Bottom line

At 20,000 emails a month, Postmark Pro is the cheapest published option at $29.50, and Mailgun's best value is Foundation at $35.00 once volume passes about 21,100 emails. Below 10,000 emails a month the decision is about the free tier, and Mailgun's 100-a-day allowance is the only one of the two that can carry live traffic.

**Sources:**
- Mailgun pricing: https://www.mailgun.com/pricing/
- Postmark pricing: https://postmarkapp.com/pricing

---

# MongoDB Atlas free tier limits: 512 MB and 100 operations per second

URL: https://actualdatabase.com/mongodb-atlas-free-tier-limits/
Verified: 2026-10-25

**Quick answer:** MongoDB Atlas's free tier is the M0 cluster: 512 MB of storage, shared RAM and shared vCPU, and up to 100 operations per second, at $0 per hour and listed as free forever. The first paid step is Flex at $0.011 per hour, which the pricing page prices at $8.00 a month in its lowest throughput band and caps at $30 a month.

- MongoDB Atlas's free tier is the M0 cluster at $0/hour: 512 MB of storage with shared RAM and shared vCPU, described on the pricing page as free forever (checked 2026-10-25).
- The free cluster is capped at 100 operations per second and at 32 MB of in-memory sort (checked 2026-10-25).
- Free clusters accept a maximum of 500 simultaneous connections (checked 2026-10-25).
- The first paid option is a Flex cluster at $0.011/hour, priced at $8.00/month for continuous use in the 0-100 operations-per-second band (checked 2026-10-25).
- Flex is advertised at up to $30/month, which matches its 400-500 operations-per-second band at $0.0411/hour, or $30.00/month (calculated 2026-10-25).
- Dedicated clusters start at $56.94/month, with M10 billed at $0.08/hour for 10-128 GB of storage, 2 GB RAM and 2 vCPU (checked 2026-10-25).

MongoDB Atlas has given away a cluster for years, and the terms have moved: shared hardware, an operations-per-second ceiling, a sort-memory cap, a connection limit buried in the documentation rather than the pricing page. This page records what the free tier actually permits on 2026-10-25 and what the first invoice looks like when a project outgrows it.

The free tier is the M0 cluster: 512 MB of storage, shared RAM and shared vCPU, at $0 per hour and listed as free forever. The first paid step is Flex at $0.011 per hour, which MongoDB prices at $8.00 a month in the lowest throughput band and caps at $30 a month.

## What the free tier actually limits

| Item | Free M0 | Flex | Dedicated M10 |
|---|---|---|---|
| Storage | 512 MB | 5 GB | 10-128 GB |
| List price | $0/hour | $0.011/hour | $0.08/hour |
| Monthly, continuous | $0 | $8.00 at base band | From $56.94 |
| Metered by | operations per second | operations per second | RAM, vCPU, storage |

The two ceilings printed in the pricing FAQ are 100 operations per second and 32 MB of sort memory. The documentation adds a third: a free cluster can hold at most 500 connections, and once the queue behind the operations-per-second limit grows, operations wait more than a second before they run.

[[ad]]

## What the numbers actually mean

Operations per second is the number to watch, because it is not requests per second. A query that scans a collection counts as many operations, so a workload with the right index stays under 100 far longer than the same traffic doing ad-hoc aggregation.

Flex is metered the same way but sold in bands. The pricing page lists five: 0-100 operations per second at $0.0110 per hour, then $0.0205, $0.0288, $0.0356 and $0.0411 as the ceiling rises by 100 each time. Run those rates over 730 hours and the bands map to $8.00, $15.00, $21.00, $26.00 and $30.00, which is where the "up to $30/month" on the Flex card comes from.

Storage is the other axis, and the first paid step is bigger than it looks. Flex includes 5 GB. The M2 free-cluster tier includes 2 GB for $9 a month; M5 includes 5 GB for $25. Against a 512 MB free cluster, M2 is a fourfold increase in space and M5 a tenfold one.

## How we measured it

Every figure here is copied from the Atlas pricing page and the free-cluster limitations page, both opened on 2026-10-25. Nothing was provisioned and no benchmark was run, so these are list prices before any commitment discount. Monthly Flex figures use the page's own assumption of continuous usage for 30 days, the same arithmetic it prints beside each band, and are labelled calculated where they are derived rather than quoted.

## What changes in practice

The gap between free and paid is where projects get hurt. M0 is not a small version of a paid cluster; it is a cluster with a throughput ceiling, a sort ceiling and a connection ceiling that never appear on a capacity chart until you hit one. The first symptom is usually latency during a deploy or a cron job, not an error.

The connection limit bites serverless deployments hardest. 500 connections sounds generous until each function invocation opens its own, and the documentation's warning about operations waiting behind the per-second queue describes a real stall rather than a slow query plan.

Flex is billed hourly, so the band you occupy during a spike is the band you pay for that hour. That makes it a reasonable home for a staging database and a poor home for a batch job you can only run at 4 a.m., where the same work on a dedicated tier bills at a known rate.

M5, at 5 GB for $25 a month, undercuts Flex's top two bands of $26 and $30 while including the same 5 GB, which makes a fixed shared cluster the cheaper home for a database whose throughput has stopped moving even as its storage grows.

## Bottom line

Under 512 MB of data and 100 operations per second, Atlas costs nothing and there is no case for paying. The moment indexes stop fitting that throughput budget, budget $9 a month for M2 and treat $30 as the Flex ceiling that decides whether you stay on shared hardware or move to a dedicated M10 at $56.94 a month.

**Sources:**
- MongoDB Atlas pricing: https://www.mongodb.com/pricing
- MongoDB Atlas free cluster limits: https://www.mongodb.com/docs/atlas/reference/free-shared-limitations/

---

# MongoDB Atlas vs Supabase free tier: 512 MB each, different meters

URL: https://actualdatabase.com/mongodb-atlas-vs-supabase-free-tier/
Verified: 2026-10-25

**Quick answer:** Both free tiers hold about half a gigabyte of database and cost $0: 512 MB on Atlas M0 and 500 MB on Supabase Free. Atlas meters throughput instead, capping free clusters at 100 operations per second and 500 connections, while Supabase leaves API requests unlimited and instead counts 50,000 monthly active users, 5 GB of egress and 2 active projects.

- Atlas M0 includes 512 MB of storage on shared RAM and shared vCPU at $0/hour, limited to 100 operations per second and 32 MB of sort memory (checked 2026-10-25).
- Supabase Free includes a 500 MB database on shared CPU with 500 MB of RAM, unlimited API requests and 50,000 monthly active users at $0/month (checked 2026-10-25).
- Supabase Free allows 2 active projects, includes 5 GB of egress plus 5 GB of cached egress and 1 GB of file storage, and pauses projects after 1 week of inactivity (checked 2026-10-25).
- Free Atlas clusters are capped at 500 simultaneous connections (checked 2026-10-25).
- The first paid step past Atlas M0 is Flex at $0.011/hour, priced at $8.00/month at the base band and capped at $30/month (checked 2026-10-25).
- The first paid step past Supabase Free is Pro at $25/month, which includes the first project and 8 GB of disk per project (checked 2026-10-25).

The two free tiers most small products start on are closer than the marketing suggests: half a gigabyte of database, shared compute, no card. What separates them is the meter each vendor puts on the plan, and that meter decides which one you outgrow first.

Atlas gives 512 MB and 100 operations per second. Supabase gives 500 MB, unlimited API requests, 50,000 monthly active users and 5 GB of egress. Both are $0, and both have a first paid step under $30 a month.

## What each free tier includes

| Item | Atlas M0 | Supabase Free |
|---|---|---|
| Database storage | 512 MB | 500 MB |
| Compute | Shared RAM, shared vCPU | Shared CPU, 500 MB RAM |
| Hard ceiling | 100 operations/second, 32 MB sort | Unlimited API requests |
| Users metered | None | 50,000 monthly active users |
| Traffic | Not listed on the free card | 5 GB egress + 5 GB cached |
| File storage | None | 1 GB |
| Project limit | Not stated | 2 active projects |
| Idle behaviour | Listed as free forever | Paused after 1 week idle |
| First paid step | Flex $0.011/hour, M2 $9/month | Pro $25/month |

Atlas's free card in the pricing table lists storage, RAM and vCPU; the throughput and sort ceilings come from the FAQ, and the 500-connection cap comes from the free-cluster documentation rather than the pricing page.

[[ad]]

## What the two vendors are actually metering

Atlas meters the database engine. Every tier from M0 to M60 is priced by storage, RAM and vCPU, and the free tier is simply the smallest allocation with a throughput ceiling of 100 operations per second. That means the free tier's failure mode is mechanical: the query that needs a collection scan, the aggregation that sorts more than 32 MB, the deploy that opens too many of its 500 connections.

Supabase meters the application around the database. API requests are unlimited on Free, which sounds like the absence of a meter until you read the rest of the card: 50,000 monthly active users, 5 GB of egress, 5 GB of cached egress, 1 GB of file storage, and a limit of 2 active projects. The database itself is 500 MB of disk on shared CPU with 500 MB of RAM, and a free project is paused after a week of inactivity.

So the units are not comparable. Atlas sells operations; Supabase sells identity and bandwidth and treats the database as a fixed allowance inside them. A read-heavy API with few users hits Atlas's wall first. A consumer app with many signups and modest queries hits Supabase's.

## How we measured it

Every figure is copied from the Atlas pricing page, the Atlas free-cluster limitations page and the Supabase pricing page, all opened on 2026-10-25. Nothing was provisioned and no benchmark was run. Where a vendor does not print a number on the page, the table says so rather than estimating it, which is why egress is blank on the Atlas side and the project limit is blank on Atlas too.

## What changes in practice

The 500 MB Supabase database is the tighter constraint for a product with content in it, because Postgres indexes and WAL live in the same allowance. The 512 MB Atlas database is equally tight, but Atlas users typically scale storage in steps: M2 is 2 GB for $9 a month, M5 is 5 GB for $25.

Supabase's inactivity pause is the one clause people discover late. A free project that gets no traffic for a week is paused, which is fine for a demo and awkward for a staging environment a client is meant to open on a Monday.

Neither free tier includes the thing that usually ends the free period: sustained traffic. Supabase's 5 GB of egress covers about 50,000 responses at 100 KB each (calculated 2026-10-25), and Atlas's 100 operations per second covers a lot of small indexed reads and very few scans. The asymmetry to check before committing is that Supabase meters egress on its free card while the Atlas free card lists no traffic allowance at all.

## Bottom line

For a prototype with a handful of users and a relational schema, Supabase Free is the better fit: 500 MB, no per-request metering and a $25 step up that includes compute credit. For a document workload with predictable small queries, Atlas M0 is cheaper to start and $9 buys four times the storage — but plan the exit at 100 operations per second, not at 512 MB.

**Sources:**
- MongoDB Atlas pricing: https://www.mongodb.com/pricing
- Supabase pricing: https://supabase.com/pricing
- MongoDB Atlas free cluster limits: https://www.mongodb.com/docs/atlas/reference/free-shared-limitations/

---

# What a Neon CU-hour is, and what an always-on database costs

URL: https://actualdatabase.com/neon-cu-hour-what-it-means/
Verified: 2026-10-25

**Quick answer:** A CU-hour is one compute unit running for one hour, and Neon bills only the CU-hours a database actually consumes because compute scales to zero when idle. Free includes 100 CU-hours per project at the default 0.25 CU, which covers about 400 active hours; an always-on database cannot fit that, because 0.25 CU running all month is about 182.5 CU-hours, or $19.35/month on Launch at $0.106 per CU-hour.

- Neon defines monthly CU-hours as average CU size multiplied by the hours the compute runs in the month, and monthly compute cost as CU-hours times the CU-hour price (checked 2026-10-25).
- The Free plan includes 100 CU-hours per project, and up to 100 projects each with its own allowance (checked 2026-10-25).
- Free plan projects default to a 0.25 CU compute size, and at that size 100 CU-hours covers about 400 hours of active compute (checked 2026-10-25).
- An always-on 0.25 CU database uses about 182.5 CU-hours a month, because 730 hours x 0.25 CU = 182.5 (checked 2026-10-25).
- At Launch's $0.106 per CU-hour, 182.5 CU-hours costs $19.35 a month; at Scale's $0.222 per CU-hour it costs $40.52 (calculated 2026-10-25).
- Neon suspends compute after 5 minutes of inactivity, and CU-hours are only consumed while a database is actively running queries (checked 2026-10-25).

Neon's bill is not a subscription with an overage; it is a metered quantity called a CU-hour, and most teams only learn what that means on their first invoice. The number is easy to reason about once you know two facts: what a CU is, and how many hours a month your database is actually awake.

A CU-hour is one compute unit running for one hour. Free includes 100 CU-hours per project at a default 0.25 CU, which is about 400 active hours. Keep the database on around the clock and the arithmetic changes: 0.25 CU x 730 hours is about 182.5 CU-hours, which is $19.35 a month on Launch.

## The CU-hour rates and what a month of compute costs

| Plan | Price per CU-hour | Always-on 0.25 CU (182.5 CU-hours) | Always-on 1 CU (730 CU-hours) |
|---|---|---|---|
| Free | 100 CU-hours included | Exceeds the allowance | Exceeds the allowance |
| Launch | $0.106 | $19.35 | $77.38 |
| Scale | $0.222 | $40.52 | $162.06 |

The pricing page states the formula in two lines: monthly CU-hours equals average CU size times the number of hours the compute runs, and monthly compute cost equals monthly CU-hours times the CU-hour price. It also prints the always-on figure itself: once scale to zero is disabled, the minimum a database can use in a month is about 182.5 CU-hours.

[[ad]]

## What the numbers actually mean

Neon sells compute the way an electricity meter sells power, and the two published rates differ by more than 2x: $0.106 per CU-hour on Launch, $0.222 on Scale. A 0.25 CU database awake all month is $19.35 on one and $40.52 on the other, for identical hardware, which is worth checking before assuming Scale is the premium plan you need.

The free allowance is the part that surprises people. 100 CU-hours is not 100 hours; it is 100 CU-hours, and at the 0.25 CU default that is 400 wall-clock hours of activity. The page's own example makes the shape concrete: a database running 0.25 CU for about 12 hours a day uses roughly 90 CU-hours a month, which fits inside Free with room to spare.

Always-on is where free ends. 0.25 CU running 730 hours is 182.5 CU-hours, nearly double the 100-hour allowance, so a free database kept permanently awake runs out somewhere past the halfway point of the month. That is a deliberate design: Neon's model assumes the database sleeps.

## How we measured it

The definition, both rates and the 182.5 CU-hour figure are quoted from Neon's pricing page on 2026-10-25; the 5-minute suspend window is from Neon's scale-to-zero documentation. The monthly totals multiply those published numbers by 730 hours, the month length the pricing page itself uses. Nothing was provisioned and no benchmark was run.

## What changes in practice

Scale to zero is the whole product decision. A demo app, a staging branch, an internal tool used eight hours a day: all of them spend most of the month suspended and pay a small multiple of their active hours. A production database behind an API that polls every few seconds never suspends, and from Neon's point of view it is always on, priced at 182.5 CU-hours minimum.

The other trap is size. CU-hours scale linearly with compute size, so a 1 CU database is four times the monthly cost of the same hours at 0.25 CU — $77.38 against $19.35 on Launch. Autoscaling up to 2 CU or 4 CU under load is billed at whichever size was active, and the pricing page's ceiling of 16 CU on Launch and 56 CU on Scale tells you how far that can go.

Storage is separate and cheaper: $0.35 per GB-month on paid plans, plus $0.20 per GB-month for history. A 100 GB database costs $35 a month in storage on either plan, which is more than an always-on 0.25 CU database and less than an always-on 1 CU one. It is also the one line that never scales to zero: a suspended database still holds its data and still bills at its rate, so the cheapest possible month on Neon is storage and nothing else.

## Bottom line

For anything that sleeps, Neon's free plan is generous — 100 CU-hours covers about 400 active hours a month at 0.25 CU. For anything always on, budget $19.35 a month on Launch before storage, and check whether you need Scale's $0.222 rate at all: the only reason to pay it is the higher compute ceiling and the larger limits, not the per-hour price.

**Sources:**
- Neon pricing: https://neon.com/pricing
- Neon docs: scale to zero: https://neon.com/docs/introduction/scale-to-zero

---

# Netlify Free plan limits in 2026: 300 credits, one member, one build at a time

URL: https://actualdatabase.com/netlify-free-plan-limits-2026/
Verified: 2026-10-25

**Quick answer:** Netlify's Free plan stops at 300 usage credits a month and allows 1 team member, 1 concurrent build, 1 concurrent agent run and 24-hour function logs. At the published rates of 15 credits per production deploy, 20 per GB of bandwidth and 10 per GB-hour of compute, 300 credits is roughly 20 deploys, 15 GB of traffic or 30 GB-hours of runtime (calculated 2026-10-25).

- Netlify Free costs $0 and ends at 300 usage credits a month, with a hard cap that cannot be exceeded or billed over (checked 2026-10-25).
- A production deploy costs 15 credits, so the Free plan covers about 20 deploys a month (calculated 2026-10-25).
- Bandwidth costs 20 credits per GB, which puts the Free plan's traffic ceiling at about 15 GB a month (calculated 2026-10-25).
- Compute costs 10 credits per GB-hour, so 300 credits equals about 30 GB-hours of function, agent or database runtime (calculated 2026-10-25).
- Web requests cost 2 credits per 10,000, so 300 credits equals about 1.5 million requests a month (calculated 2026-10-25).
- Free includes 1 team member, 1 concurrent build, 1 concurrent agent run, 1 preview server and 24 hours of function and edge logs (checked 2026-10-25).

The Netlify Free plan is not a bundle of build minutes and gigabytes in 2026; it is a single budget of 300 usage credits a month. Every billable resource draws from that one number, so the answer to "how much can I run for free" is a division problem rather than a list of allowances. At the published rates, 300 credits buys about 20 production deploys, about 15 GB of bandwidth, about 30 GB-hours of compute or about 1.5 million web requests. The table below keeps those conversions in one place so you can see which limit your project will hit first.

## What the Free plan covers

| Resource | Free plan | Metered cost |
|---|---|---|
| Usage credits | 300 per month, hard cap | $5 per 500 or $10 per 1,500 (paid plans) |
| Production deploys | 15 credits each, about 20 a month | $0.10 each |
| Bandwidth | 20 credits per GB, about 15 GB a month | $0.13 per GB |
| Compute | 10 credits per GB-hour, about 30 GB-hours | $0.07 per GB-hour |
| Web requests | 2 credits per 10k, about 1.5 million | $0.01 per 10,000 |
| Team members | 1 | unlimited on Pro at $20 a month |
| Concurrent builds | 1 | 3 on Pro, extra builds $40 each |
| Function and edge logs | 24 hours | 7-day observability on Pro |
| Deploy previews | unlimited | unlimited on every plan |

The dollar column is Netlify's own conversion, taken from the credits-to-USD calculator that prices everything at the Pro credit-pack rate of $10 per 1,500 credits. The per-credit figure that comes out of that is $0.00667, and it is the number to use whenever you compare a Netlify line item with a competitor's list price.

[[ad]]

## What the 300-credit cap actually means

The tightest limit on Free is rarely bandwidth. It is deploys: 15 credits each means a repository that pushes to production 20 times in a month consumes the entire allowance and has nothing left for traffic. A team that ships daily crosses that line in under three weeks, and the pause that follows is account-wide — Netlify pauses every project on the account, not just the one that ran out.

The second surprise is that compute and bandwidth share the budget. Functions, Agent Runners and Netlify Database compute all bill at 10 credits per GB-hour, so a function that idles warm for a month at 0.5 GB will burn through credits even when nobody visits the site. There is no separate functions allowance to fall back on.

Free also cannot be topped up. Netlify sells extra credits as add-ons, but the pack table attaches 500 credits to Personal and 1,500 to Pro, and auto-recharge is off by default on paid plans. On Free the only lever is waiting for the next billing cycle.

## How we measured it

Both Netlify pages were fetched on 2026-10-25: the rendered pricing page at netlify.com/pricing and its markdown twin at netlify.com/pricing.md, which carries the credit-cost table and the credits-to-USD calculator in plain text. No benchmark, load test or billing experiment was run — every figure here is either copied from those two pages or divided out of them, and every derived number is marked as calculated. Where the page publishes no rate, such as the credit cost of a form submission, this article says so instead of guessing.

## What changes in practice

The 24-hour log window is the limit that hurts after the fact. Observability on Free is not part of the plan, Personal adds 1 day and Pro adds 7 days, so an incident that starts on a Friday is effectively undebuggable by Monday on Free. If you run anything with a login form, budget for Personal just to keep logs.

One concurrent build is a queue, not a failure. Two pushes in a row serialize, and a monorepo with three apps builds them one after another. Pro raises the count to 3 and then charges $40 for each additional concurrent build, which is a per-slot fee rather than a per-minute fee — the economics only make sense if slow builds are actually blocking deploys.

One member is the quiet limit on agency work. Git Contributors are free and unlimited on public repositories, and Reviewers never consume a seat, so the single-member cap mostly blocks a second developer who needs deploy rights. That is when a $20 Pro plan shows up on a card, not a bandwidth bill.

## Bottom line

Free is a real $0 tier for a site that deploys weekly and serves under about 15 GB a month, and it fails predictably once either of those numbers moves. When it does, skip Personal: $20 for 3,000 credits works out to $0.00667 per credit against Personal's $0.009, so Pro is about 26 percent cheaper per credit and throws in the extra seats (calculated 2026-10-25).

**Sources:**
- Netlify pricing: https://www.netlify.com/pricing/
- Netlify pricing, credit costs and credit packs (markdown version): https://www.netlify.com/pricing.md

---

# Netlify Pro in 2026: $20 a month, 3,000 credits and unlimited members

URL: https://actualdatabase.com/netlify-pro-what-it-includes/
Verified: 2026-10-25

**Quick answer:** Netlify Pro costs $20 a month and includes 3,000 usage credits, unlimited team members, 3 concurrent builds, private organization repositories, shared environment variables, 30-day analytics and 7-day observability. The limit that forces most upgrades is the seat: Free and Personal are individual plans with 1 member, and Pro is the first Team tier.

- Netlify Pro costs $20 a month and includes 3,000 usage credits, which works out to $0.00667 per credit (checked 2026-10-25).
- Pro is the first plan with unlimited members, while Free and Personal are individual plans limited to 1 member (checked 2026-10-25).
- Pro credits come in tiers of 3,000, 5,000, 10,000, 15,000 and 20,000 per month, with extra credits sold at $10 per 1,500 (checked 2026-10-25).
- Pro raises concurrent builds from 1 to 3, and every additional concurrent build after that costs $40 (checked 2026-10-25).
- Pro adds 7-day observability and 30 builds retained, against 1 day of observability and 5 builds on Free (checked 2026-10-25).
- 3,000 credits is 10 times the Free allowance and covers about 150 GB of bandwidth, 60 production deploys or 300 GB-hours of compute (calculated 2026-10-25).

Netlify Pro is $20 a month for 3,000 usage credits and, more importantly, for the right to add a second person to the account. The plan sits between Personal at $9, which is still an individual plan with 1 member, and Enterprise at custom pricing. Pro itself has not become complicated: unlimited members, 3 concurrent builds, private organization repositories, shared environment variables, 30-day analytics and 7-day observability, all on the same credit meter that runs the Free tier. What surprises people is which of those lines is the one they actually needed.

## What Pro includes

| Line item | Free | Pro at $20 a month |
|---|---|---|
| Usage credits | 300 per month, hard cap | 3,000 per month, tiers to 20,000 |
| Members | 1 | unlimited |
| Concurrent builds | 1 | 3, then $40 per extra slot |
| Concurrent agent runs | 1 | 10 |
| Preview servers | 1 | 1, then $15 each |
| Observability | not included | 7 days |
| Builds retained | 5 | 30 |
| Private org repositories | no | yes |
| Shared environment variables | no | yes |
| Analytics and metrics | basic | 30 days |
| Support | community | priority email |

Personal at $9 sits in the same table with 1,000 credits, 1 member, smart secret detection, 1-day observability and priority email support. It is a credit top-up with better logs, not a team plan, and the page labels its audience "Individual" the same way it labels Free.

[[ad]]

## What the $20 actually means

The seat count is the limit that forces this upgrade. Deploy permissions are tied to account membership, and Free and Personal both cap membership at 1, so the moment a second developer needs to push to production the choices are Pro or sharing a login. At $20 for unlimited members, the marginal cost of the fifth developer is zero, which is why the price discussion usually ends quickly once a team is involved.

Credits alone rarely justify Pro. Personal's 1,000 credits already covers 50 GB of bandwidth or 66 deploys, and a site that fits inside 1,000 credits has no reason to pay for 3,000. Where the credit tiers matter is at the top end: 3,000, 5,000, 10,000, 15,000 and 20,000 credits a month are published as steps, so a heavy month lands on a bigger tier rather than on a per-unit overage bill.

The extra concurrent build is worth pricing separately. Pro includes 3 and charges $40 for each additional slot, so a monorepo that wants 6 parallel builds pays $20 for the plan plus $120 for slots. That is a fixed monthly fee per slot, not a usage charge, and it is the most expensive thing on the Pro line for anyone with many small apps.

## How we measured it

The plan table and the credit-cost table were fetched on 2026-10-25 from netlify.com/pricing and from the plain markdown version of the same page at netlify.com/pricing.md. Nothing was benchmarked and no account was upgraded; every figure is copied from those two pages, and every number that involved division — credits into gigabytes, dollars into credits — is labelled as calculated.

## What changes in practice

Pro is a Team plan, so the things that change are mostly collaborative. Private organization repositories mean a private GitHub org can deploy without a public mirror, shared environment variables mean secrets stop being copy-pasted per site, and 30-day analytics mean a monthly review does not depend on someone exporting CSVs on the first of the month.

The credit behaviour does not change at all. Pro still has a hard monthly cap on each tier, and hitting it still pauses the whole account rather than silently billing an overage, so the tier you pick is a budget you have to size correctly. The difference from Free is only that the pause is further away and the notices arrive earlier.

Observability is the underrated line. Going from no window on Free to 7 days on Pro is what makes a production incident diagnosable after the weekend, and it is the reason some single-person projects buy Pro even though they never add a member.

## Bottom line

Buy Pro at $20 when a second person needs deploy rights or when 1,000 credits is not enough — both are seat and volume problems that Personal cannot solve at any price. If neither applies, Personal at $9 with 1,000 credits is the cheaper answer, and $0.009 per credit versus Pro's $0.00667 only matters once monthly spend passes the Personal ceiling (calculated 2026-10-25).

**Sources:**
- Netlify pricing: https://www.netlify.com/pricing/
- Netlify pricing, credit costs and credit packs (markdown version): https://www.netlify.com/pricing.md

---

# 100 GB of static traffic in 2026: $20 on Netlify against $14.25 on Render

URL: https://actualdatabase.com/netlify-vs-render-static-site-bandwidth/
Verified: 2026-10-25

**Quick answer:** 100 GB of monthly traffic costs $20 on Netlify, because bandwidth is 20 credits per GB and 2,000 credits requires the Pro plan, against $14.25 on Render, which includes 5 GB and charges $0.15 per GB after it. Netlify's free tier ends at 15 GB and pauses every project there, while Render's free inclusion ends at 5 GB and keeps serving on a metered bill.

- 100 GB of monthly static traffic costs $20 on Netlify, because bandwidth is metered at 20 credits per GB and 2,000 credits only fits the Pro plan (calculated 2026-10-25).
- The same 100 GB costs $14.25 on Render's $0 Hobby workspace: 5 GB included, then 95 GB at $0.15 per GB (calculated 2026-10-25).
- Netlify's free tier ends at 15 GB, exactly the 300-credit monthly cap, and hitting it pauses every project on the account until the next cycle (checked 2026-10-25).
- Render's free inclusion ends at 5 GB, but traffic past it is billed at $0.15 per GB rather than blocked (checked 2026-10-25).
- At 50 GB, Netlify needs its $9 Personal plan for 1,000 credits while Render bills $6.75, and at 500 GB the position reverses: about 10,000 Netlify credits against $74.25 on Render (calculated 2026-10-25).
- Both platforms count all workspace traffic, so a static site's bandwidth competes with API responses and preview deployments for the same allowance (checked 2026-10-25).

A static site has one cost that scales with success and almost nothing else: outbound bandwidth. Two platforms that both host static sites for free price that gigabyte very differently, and the differences only appear once a site is popular. On Netlify, bandwidth costs 20 credits per GB against a hard 300-credit free cap, so 100 GB of traffic needs the $20 Pro plan. On Render it costs $0.15 per GB after 5 GB included, so the same 100 GB bills $14.25 on the free Hobby workspace. Cheaper is not the whole story, because one of these platforms stops serving at the limit and the other sends an invoice.

## 100 GB on each platform

| Monthly traffic | Netlify | Render |
|---|---|---|
| 5 GB | 100 credits, $0 on Free | included, $0 |
| 15 GB | 300 credits, exactly the Free cap | $1.50 |
| 50 GB | 1,000 credits, $9 Personal | $6.75 |
| 100 GB | 2,000 credits, $20 Pro | $14.25 |
| 500 GB | 10,000 credits, a larger Pro tier | $74.25 |

The Netlify column is not a price list, it is a plan ladder. Credits are sold in blocks, so the bill jumps from $0 to $9 to $20 as traffic crosses 15 GB and then 50 GB, and the credits paid for but unused at each tier are not refunded.

[[ad]]

## What the credit ladder actually means

Netlify replaced per-resource limits with credits, and bandwidth is the most expensive of its meters at 20 credits per GB. The Free plan's 300 credits therefore buy exactly 15 GB of traffic and nothing else — no deploys, no function compute, no web requests. A site sitting at 15 GB of traffic on the Free plan has used its entire allowance on bandwidth alone, which is the point where the free tier ends.

What happens at that point differs more than the prices do. Netlify's Free plan has hard monthly limits that cannot be exceeded, and crossing a project's limit pauses all projects on the account until the next billing cycle. Render's Hobby plan includes 5 GB and then bills $0.15 per GB with no cap, so the site keeps serving and the owner discovers the overage later. One model fails closed, the other fails open and expensive.

Render's 5 GB is smaller than Netlify's 15 GB, but it is a billing threshold rather than a wall. For a site with steady traffic the second model is easier to live with; for a site that might be linked from somewhere large, the first model is a hard stop that costs nothing.

## How we measured it

Both pricing pages were fetched on 2026-10-25. Netlify's page supplies the 20-credits-per-GB rate, the 300-credit Free cap, the 1,000-credit Personal plan at $9 and the 3,000-credit Pro plan at $20; Render's supplies the 5 GB Hobby allowance and the $0.15 per GB overage. Every dollar figure in the table is arithmetic on those rates at the stated traffic levels, and no site was deployed or measured on either platform.

## What changes in practice

The credits are shared across every meter, so bandwidth competes with everything else the project does. A Netlify site that uses 10 GB of traffic has spent 200 of its 300 credits and has 5 GB left for deploys, function compute and web requests — 15 production deploys at 15 credits each would exhaust the remainder before the traffic did. The true free ceiling is lower than 15 GB for any project that also deploys.

Render's allowance is shared too, but only with traffic. Build minutes are metered separately at 500 a month on Hobby, and compute is priced per instance, so the 5 GB is genuinely 5 GB of traffic. That makes Render's free tier easier to reason about even though its number is smaller.

Large traffic inverts the comparison. At 500 GB, Render's overage alone is $74.25, while Netlify needs one of the 10,000-credit Pro tiers, about $66.67 at the page's own credit pack rate of $10 per 1,500 credits (calculated 2026-10-25). Neither is cheap, and at that volume a CDN in front of an object store is the conversation worth having.

## Bottom line

For a static site under 100 GB a month, Render is the cheaper host at $14.25 against $20, and it is the safer one if traffic is unpredictable because it bills instead of pausing. Choose Netlify only if the 15 GB free tier covers the site, since past that point its credit ladder costs more per gigabyte than Render's flat $0.15 (calculated 2026-10-25).

**Sources:**
- Netlify pricing: https://www.netlify.com/pricing/
- Render pricing: https://render.com/pricing

---

# Two newsletters a month to 5,000 subscribers: $29 on Buttondown, calculator-only on Mailchimp

URL: https://actualdatabase.com/newsletter-email-cost-5000-subscribers/
Verified: 2026-10-25

**Quick answer:** Buttondown Standard costs $29/month for up to 5,000 subscribers and its published pricing assumes up to one send a day to the whole list, so two sends a month is $29 — $14.50 per send. Mailchimp does not publish a 5,000-contact price in its page source: its Essentials plan starts at $13/month for 500 contacts and its published ceiling is $385/month at 50,000 contacts, so the 5,000-contact figure has to come from its calculator.

- Buttondown Standard costs $29/month for up to 5,000 subscribers, with Basic at $9 for 1,000 and Professional at $79 for 10,000 (checked 2026-10-25).
- Buttondown's prices assume at most one email a day to the entire subscriber base, and it charges for active subscribers rather than stored contacts (checked 2026-10-25).
- Two sends to 5,000 subscribers is 10,000 emails a month, or $14.50 per send on Buttondown Standard (calculated 2026-10-25).
- Mailchimp's Essentials plan starts at $13/month for up to 500 contacts and is published at $385/month for 50,000 contacts (checked 2026-10-25).
- Mailchimp's Standard plan starts at $20/month for up to 500 contacts and is published at $800/month for 100,000 contacts (checked 2026-10-25).
- At 5,000 contacts, Mailchimp's Essentials monthly send limit is 10 times the contact count, or 50,000 sends (calculated 2026-10-25).

A newsletter list of 5,000 is the point where free tiers end and pricing pages get vague. One of the two vendors here states the number outright; the other sells it through a slider.

Buttondown Standard is $29 a month for up to 5,000 subscribers, and its published model assumes at most one send a day, so two sends cost $29 — $14.50 per send. Mailchimp publishes $13 a month for its first 500 contacts and $385 a month at 50,000; the band that contains 5,000 is not in the page source at all.

## What 5,000 subscribers cost

| Item | Buttondown Standard | Mailchimp Essentials |
|---|---|---|
| Published plan price | $29/month | $13/month at 500 contacts |
| Published ceiling | 10,000 subscribers at $79 | 50,000 contacts at $385 |
| Price at 5,000 | $29/month | Not published in page source |
| Sends allowed | Up to one a day to the list | 10 x contacts = 50,000/month |
| Two sends to 5,000 | 10,000 emails, no extra charge | 10,000 sends of 50,000 |
| Free plan ceiling | 100 subscribers | 250 contacts, 500 sends |

Mailchimp's send allowance is a multiplier, not a fixed number: Essentials sets the monthly send limit at 10 times the contact count, Standard at 12 times and Premium at 15 times. Mailchimp's own definition is that one email sent to 500 contacts equals 500 sends.

[[ad]]

## What the numbers actually mean

Buttondown's price is per subscriber tier and nothing else. The ladder runs $9 for 1,000, $29 for 5,000, $79 for 10,000 and $139 for 20,000, and it charges for active subscribers rather than everyone who ever signed up. Two sends a month do not move the meter, because the plan assumes up to one a day.

Mailchimp's price is per contact band, and the band is the whole product. The published anchors bracket 5,000 tightly: $13 a month buys 500 contacts and $385 a month buys 50,000 on the same plan. A price anywhere between those two is plausible for 5,000 contacts, which is exactly why the calculator exists and why this page will not print a number it cannot source.

The send allowance matters less than list owners expect. At 5,000 contacts an Essentials account can send 50,000 emails a month, so two campaigns use a fifth of it and ten campaigns still fit. Frequency is not what costs money on either platform; list size is.

## How we measured it

Buttondown's tier prices were read from the plan data on its pricing page, and its send assumption from the FAQ on that page, on 2026-10-25. Mailchimp's base prices, contact ceilings and send-limit multipliers come from its pricing page and its plans help article, both opened the same day; the 5,000-contact band price was searched for across the pricing page, the marketing pricing page and the plan comparison table, and does not exist in static form. No account was created and no email was sent.

## What changes in practice

The first difference is what you can budget. $29 is a fixed number you can put in a spreadsheet. A Mailchimp bill at 5,000 contacts is a function of a contact band that changes the moment 400 people subscribe, and Mailchimp's help article is explicit that the Essentials price increases with the contact total.

The second is overshoot. Buttondown's tiers are ceilings rather than metered bands, so 5,100 subscribers means moving to the $79 Professional tier for 10,000 — a 172% jump for 100 extra people. Mailchimp's bands move in smaller steps but apply to every contact, so growth is charged immediately rather than at the next boundary.

The third is what happens when nobody opens the email. Buttondown charges for active subscribers; Mailchimp counts stored contacts, so a list that has gone cold still bills at full band. That is the single line most likely to explain a large difference between two quotes for the same 5,000 people.

A final check is the cost of crossing a tier boundary. Buttondown's next tier is $79 for 10,000 subscribers, 2.7 times the $29 Standard price for twice the list (calculated 2026-10-25). Mailchimp's bands are narrower, but they apply to every contact, so growth past a band boundary reprices the whole account rather than just the new subscribers.

## Bottom line

At 5,000 subscribers and two sends a month, Buttondown's $29 is the number to budget: it includes the volume, charges nothing per send, and sits $50 below the next tier. Mailchimp is only comparable once you run its calculator — and because its send allowance is ten times the contact count, a list that mails twice a month is being charged for audience size, not for sending.

**Sources:**
- Buttondown pricing: https://buttondown.com/pricing
- Mailchimp pricing plans: https://mailchimp.com/pricing/
- Mailchimp help: about pricing plans: https://mailchimp.com/help/about-mailchimp-pricing-plans/

---

# LLM API Prices per Million Tokens: 10 Models Side by Side

URL: https://actualdatabase.com/openrouter-llm-prices-per-million-tokens/
Verified: 2026-10-25

**Quick answer:** On OpenRouter's published rates from 2026-10-25, Claude Sonnet 4.5 costs $3.00 per million input tokens and $15.00 per million output, GPT-5 costs $1.25 and $10.00, Gemini 2.5 Flash costs $0.30 and $2.50, and Llama 3.3 70B costs $0.10 and $0.32. Output tokens cost 3x to 8x more than input tokens on every model in this set.

- GPT-5: $1.25 per million input tokens, $10.00 per million output, 400,000 token context (checked 2026-10-25 via OpenRouter's model API).
- Claude Sonnet 4.5: $3.00 input, $15.00 output, 1,000,000 token context (checked 2026-10-25).
- Gemini 2.5 Flash: $0.30 input, $2.50 output, 1,048,576 token context (checked 2026-10-25).
- DeepSeek Chat: $0.257 input, $1.029 output, 163,840 token context (checked 2026-10-25).
- Llama 3.3 70B on OpenRouter: $0.10 input, $0.32 output, 131,072 token context (checked 2026-10-25).
- OpenRouter's model API listed 440 models with a positive price on 2026-10-25.

Model prices change faster than any other number on this site, which is exactly why a dated table beats a blog post from last spring.

These rates were pulled from OpenRouter's public model API on 2026-10-25. The API returns per-token prices for every model it routes; the table below converts them to dollars per million tokens, which is the unit people actually reason in.

## Ten models, priced per million tokens

| Model | Input / 1M | Output / 1M | Output ÷ input | Context |
|---|---|---|---|---|
| Llama 3.3 70B | $0.10 | $0.32 | 3.2x | 131,072 |
| GPT-4o mini | $0.15 | $0.60 | 4.0x | 128,000 |
| DeepSeek Chat | $0.257 | $1.029 | 4.0x | 163,840 |
| GPT-5 mini | $0.25 | $2.00 | 8.0x | 400,000 |
| Gemini 2.5 Flash | $0.30 | $2.50 | 8.3x | 1,048,576 |
| Claude Haiku 4.5 | $1.00 | $5.00 | 5.0x | 200,000 |
| GPT-5 | $1.25 | $10.00 | 8.0x | 400,000 |
| Gemini 2.5 Pro | $1.25 | $10.00 | 8.0x | 1,048,576 |
| Mistral Large | $2.00 | $6.00 | 3.0x | 128,000 |
| Claude Sonnet 4.5 | $3.00 | $15.00 | 5.0x | 1,000,000 |

[[ad]]

## What a real workload costs

Prices per million tokens are hard to feel. A support assistant that handles 20,000 conversations a month, each with 2,000 input tokens of context and 400 tokens of answer, moves 40 million input and 8 million output tokens.

| Model | Monthly cost for that workload |
|---|---|
| Llama 3.3 70B | $6.56 |
| Gemini 2.5 Flash | $32.00 |
| GPT-5 mini | $26.00 |
| GPT-5 | $130.00 |
| Claude Sonnet 4.5 | $240.00 |

The spread between the cheapest and most expensive option is 36x for identical token counts. That is the whole decision: on most chat products, the model that is 5% better does not justify being 10x more expensive, and on code or long-document work the cheap model fails often enough that retries erase the saving.

## How we measured it

We requested OpenRouter's `/api/v1/models` endpoint on 2026-10-25, parsed the `pricing.prompt` and `pricing.completion` fields as dollars per token, and multiplied by one million. Models with a zero or missing price were excluded, leaving 440 priced models. The ten in the table were chosen to cover one model per vendor at both a frontier and a budget tier; nothing was measured by running prompts.

## What changes in practice

The output-to-input ratio varies more than the headline price. Mistral Large charges 3x for output; GPT-5 and Gemini charge 8x. A workload that reads long documents and writes short answers is far cheaper than its input price suggests, while a workload that generates long reports pays almost entirely at the output rate. Comparing models by input price alone gets the ranking wrong for exactly those cases.

Caching and batch discounts can move a bill by half again, and neither is visible in a per-token table — you have to read each provider's own pricing page for those. The numbers here are the list rate before any of that.

## Bottom line

For high-volume chat where occasional wrong answers are survivable, the $0.10–$0.30 input tier is 8x to 12x cheaper than the frontier tier and good enough. For anything that has to be right the first time — code, extraction, legal text — the $1.25–$3.00 tier earns its price back in retries avoided. The output rate, not the input rate, is what to model your bill on.

**Sources:**
- OpenRouter models API: https://openrouter.ai/api/v1/models

---

# PlanetScale HA vs single node: what 3× the price buys

URL: https://actualdatabase.com/planetscale-ha-vs-single-node-cost/
Verified: 2026-10-25

**Quick answer:** PlanetScale charges exactly 3× the single-node rate for a high-availability cluster, because it runs three nodes: $15/month instead of $5/month at PS-5, and $4,529/month instead of $1,510/month at PS-2560. You are buying two extra replicas across three availability zones, not a different service tier.

- PlanetScale Postgres HA is $15/month at PS-5 against $5/month for the single node, a 3.0x multiplier (checked 2026-10-25).
- The multiplier holds at every size: PS-40 is $83/month for 3 nodes against $28/month for one on arm64, and PS-2560 is $4,529 against $1,510 (checked 2026-10-25).
- HA is described as 1 primary plus 2 replicas across 3 availability zones, which is why the price is three compute nodes (checked 2026-10-25).
- The absolute cost of HA at the entry SKU is $180/year against $60/year for a single node (calculated 2026-10-25).
- At PS-160 the HA premium is $190/month on arm64 ($286 minus $96), and at PS-2560 it is $3,019/month (calculated 2026-10-25).
- Storage, backups and egress are billed separately from both figures, so the multiplier applies to compute only (checked 2026-10-25).

PlanetScale publishes two columns for the same database size, and the second one is always exactly three times the first. That is not a fee and not a support tier — it is one primary plus two replicas across three availability zones. Whether the multiplier is worth paying depends on what an hour of downtime costs you, and at the entry SKU the answer is easier than it looks.

## The same database at 1× and 3×

Monthly USD, AWS us-east-1, arm64 unless stated.

| SKU | Single node | 3-node HA | Difference per month | Multiplier |
|---|---|---|---|---|
| PS-5 (512 MiB) | $5 | $15 | $10 | 3.0× |
| PS-20 (2 GiB) | $17 | $50 | $33 | 2.9× |
| PS-80 (8 GiB) | $50 | $148 | $98 | 3.0× |
| PS-160 (16 GiB) | $96 | $286 | $190 | 3.0× |
| PS-640 (64 GiB) | $379 | $1,135 | $756 | 3.0× |
| PS-2560 (256 GiB) | $1,510 | $4,529 | $3,019 | 3.0× |

The PS-20 row is the only one that is not a clean 3.0×, because the HA price rounds to $50 while the single node is $17 (ratio 2.94×); every other tier divides exactly.

[[ad]]

## What the extra money actually buys

Two more copies of the same compute, in two more availability zones. PlanetScale states the topology plainly: high availability is one primary and two replicas across three AZs, and the catalog bills it as a 3-node SKU. Nothing about the primary changes — same vCPU, same RAM, same connection limit — so the purchase is availability, not performance.

That has a practical consequence: HA does not make slow queries faster. A workload that is CPU-bound at PS-40 will be CPU-bound at PS-40 HA, and pay $166/month more for the privilege (calculated 2026-10-25).

The absolute numbers matter more than the multiplier. At PS-5 the premium is $10 a month — $120 a year — which is less than a single hour of engineering time in most companies. At PS-2560 it is $3,019 a month, and the same reasoning stops applying: at that size a serious incident review usually happens before the invoice does.

## How we measured it

We read PlanetScale's generated price catalog on 2026-10-25 (generated at 14:49:57Z) and divided each HA row by its single-node counterpart. Both columns come from the same table, so the ratio is exact rather than estimated: three nodes at the published per-node rate. No cluster was provisioned for this page.

## What changes in practice

The multiplier applies to compute only. Storage, backup storage, egress and an optional dedicated PgBouncer are billed on top of whichever column you pick, so an HA cluster is not three times the total bill — it is three times the compute line, plus the same storage.

For a comparison against a bundled plan, the fair entry point is Supabase Pro at $25/month, which includes a small database, storage and backups but no automatic three-zone replica set. A PlanetScale PS-5 HA cluster at $15/month is cheaper on compute and dearer once storage and egress are added, which is why the two products are rarely priced by their headline number alone.

One more lever: region and architecture. The catalog is priced for AWS us-east-1, and moving the same SKU to x86-64 costs 30% more at PS-10 and 23.6% more at PS-2560. Choosing arm64 is often a bigger saving than choosing single node over HA at small sizes — but it is also the choice that cannot be reversed in a hurry.

## Bottom line

Take HA from the start at PS-5 or PS-10: $15 and $30 a month is cheap insurance against a zone failure, and the $120–$260 annual premium is smaller than most incident reviews. Above PS-320 the decision reverses — $699 and up per month for redundancy is a budget line, and it should be justified by a measured recovery-time requirement rather than by default.

**Sources:**
- PlanetScale pricing catalog (machine-readable): https://planetscale.com/pricing.md
- PlanetScale pricing: https://planetscale.com/pricing
- Supabase pricing (reference point): https://supabase.com/pricing

---

# PlanetScale Postgres pricing in 2026: $5 single node to $4,529 HA

URL: https://actualdatabase.com/planetscale-postgres-pricing-per-month/
Verified: 2026-10-25

**Quick answer:** A single-node PlanetScale Postgres cluster starts at $5/month for 512 MiB of RAM, and the same SKU with high availability (one primary plus two replicas across three availability zones) costs $15/month — exactly three times the single-node rate at every tier up to PS-2560 at $4,529/month.

- PlanetScale Postgres PS-5 single node (1/16 vCPU, 512 MiB) costs $5/month on both arm64 and x86-64 (checked 2026-10-25).
- The same PS-5 SKU in 3-node HA costs $15/month, which is exactly 3x the single-node price (checked 2026-10-25).
- PS-20 single node is $17/month on arm64 and $20/month on x86-64; in HA the two architectures are $50 and $59 (checked 2026-10-25).
- The largest SKU, PS-2560 (32 vCPU, 256 GiB), is $1,510/month on arm64 and $1,867/month on x86-64 single node; HA is $4,529 and $5,599 respectively (checked 2026-10-25).
- Storage, backup storage, egress and optional dedicated PgBouncer are billed separately from the compute SKU rate (checked 2026-10-25).
- Prices in the catalog are quoted for AWS us-east-1 and change by region and cloud provider (checked 2026-10-25).

PlanetScale sells Postgres in fixed sizes, and the size table is public in a machine-readable file, so the arithmetic needs no calculator: a single-node cluster starts at $5 a month, the same SKU with three nodes costs $15, and the price ladder runs to $4,529 for 32 vCPU with high availability. What the SKU price does not include is storage, backups and egress — and that is where a monthly bill gets decided.

## The Postgres price ladder

All prices are monthly, in USD, from PlanetScale's published catalog for AWS us-east-1.

| SKU | vCPU / RAM | Single node, arm64 | Single node, x86-64 | 3-node HA, arm64 | 3-node HA, x86-64 |
|---|---|---|---|---|---|
| PS-5 | 1/16 · 512 MiB | $5 | $5 | $15 | $15 |
| PS-10 | 1/8 · 1 GiB | $10 | $13 | $30 | $39 |
| PS-20 | 1/4 · 2 GiB | $17 | $20 | $50 | $59 |
| PS-40 | 1/2 · 4 GiB | $28 | $33 | $83 | $99 |
| PS-80 | 1 · 8 GiB | $50 | $60 | $148 | $179 |
| PS-160 | 2 · 16 GiB | $96 | $117 | $286 | $349 |
| PS-320 | 4 · 32 GiB | $190 | $233 | $570 | $699 |
| PS-640 | 8 · 64 GiB | $379 | $467 | $1,135 | $1,399 |
| PS-1280 | 16 · 128 GiB | $755 | $933 | $2,265 | $2,799 |
| PS-2560 | 32 · 256 GiB | $1,510 | $1,867 | $4,529 | $5,599 |

[[ad]]

## What the numbers actually mean

High availability is priced as three machines, because it is three machines: one primary and two replicas spread across three availability zones. Every row above reproduces that exactly — $5 becomes $15, $755 becomes $2,265, $1,510 becomes $4,529 — so there is no discount for buying the redundant topology, and no hidden premium either (calculated 2026-10-25).

The architecture column is the other lever. arm64 and x86-64 cost the same at PS-5, then diverge: 30% more for x86-64 at PS-10 ($13 against $10) and 23.6% more at PS-2560 ($1,867 against $1,510). If a workload is not architecture-bound, the arm64 column is simply cheaper at every size above the entry SKU.

## How we measured it

We read PlanetScale's own catalog file on 2026-10-25 at 14:49:57Z, which is generated from the same price source as the pricing page and states its basis: AWS us-east-1. We did not run a database, and we did not query the read-only Postgres endpoint PlanetScale publishes for live prices, so figures are the published list rates rather than a regional quote.

## What changes in practice

The SKU rate is compute only. The pricing page lists the rest as separate factors: storage and provisioned IOPS, backup storage, network egress, a dedicated PgBouncer, and extra development branch hours. That means two clusters with the same $50 line item can bill very differently, and the difference is usually storage and egress rather than CPU.

Region matters too. The catalog's own caveat is that prices are based on AWS us-east-1 and vary by cloud and region, so a workload placed in Europe or on GCP should be priced in the console rather than from this table. Metal instances invert one of the assumptions: local NVMe storage is included in the SKU rate instead of billed as EBS.

For capacity planning, the useful reading is price per GiB of RAM: $9.77 per GiB at PS-5 single node, $5.90 at PS-2560 on arm64 (calculated 2026-10-25). Small clusters are expensive per gigabyte, which is the normal shape of a fixed-size ladder.

## Bottom line

If you need a Postgres database that is cheap to start and predictable to scale, PS-5 at $5/month or PS-10 at $10/month on arm64 covers development and low-traffic production; budget for the 3× step to HA at $15 and $30 the moment the data matters, and price storage and egress separately before comparing the total against a bundled plan.

**Sources:**
- PlanetScale pricing catalog (machine-readable): https://planetscale.com/pricing.md
- PlanetScale pricing: https://planetscale.com/pricing

---

# What 100 GB of Postgres storage costs per month: Supabase, Neon and Turso

URL: https://actualdatabase.com/postgres-storage-cost-100gb/
Verified: 2026-10-25

**Quick answer:** 100 GB of database storage costs $11.50/month on Supabase Pro at $0.125 per GB above the included 8 GB, $35.00/month on Neon Launch at $0.35 per GB-month with nothing included, and $68.25/month on Turso Developer at $0.75 per GB above its 9 GB. Turso Scaler is cheaper than Turso Developer at this size: $38.00 for the storage line, $62.92 including its plan fee.

- Supabase Pro includes 8 GB of disk per project and charges $0.125 per GB beyond it (checked 2026-10-25).
- Neon charges $0.35 per GB-month of database storage on Launch with no included allowance, plus $0.20 per GB-month for history storage (checked 2026-10-25).
- Turso Developer includes 9 GB and charges $0.75 per GB beyond it; Turso Scaler includes 24 GB and charges $0.50 per GB (checked 2026-10-25).
- 100 GB of storage costs $11.50 on Supabase Pro, $35.00 on Neon Launch, $68.25 on Turso Developer and $38.00 on Turso Scaler (calculated 2026-10-25).
- Adding the plan fee, 100 GB costs $36.50/month on Supabase Pro, $73.24/month on Turso Developer and $62.92/month on Turso Scaler (calculated 2026-10-25).
- 100 GB does not fit any free tier: Supabase Free includes 500 MB, Neon Free 1 GB per project and Turso Free 5 GB (checked 2026-10-25).

A hundred gigabytes is the size where a database stops being a demo and the storage line starts to have an opinion. It is also the size where included allowances still matter, because every one of these three vendors bundles some storage before the per-gigabyte rate applies.

At 100 GB the storage line is $11.50 on Supabase Pro, $35.00 on Neon Launch, $68.25 on Turso Developer and $38.00 on Turso Scaler. Add each plan's own fee and the totals separate further.

## The storage line at 100 GB

| Provider and plan | Included | Rate per GB | Storage for 100 GB | Monthly total |
|---|---|---|---|---|
| Supabase Pro | 8 GB per project | $0.125 | $11.50 | $36.50 |
| Neon Launch | None | $0.35/GB-month | $35.00 | $35.00 + compute |
| Turso Developer | 9 GB | $0.75 | $68.25 | $73.24 |
| Turso Scaler | 24 GB | $0.50 | $38.00 | $62.92 |

The arithmetic is (100 GB minus the included allowance) times the rate, then the plan fee added back. Neon has no plan fee on Launch, so its total is the storage line plus whatever compute the database uses.

[[ad]]

## What the 100 GB numbers actually mean

The ordering here is not the ordering at 1 TB. At 100 GB, Turso Scaler is the second cheapest option and Turso Developer is the most expensive, purely because Scaler's 24 GB allowance and $0.50 rate beat Developer's 9 GB and $0.75. A team on Developer paying $73.24 for 100 GB is paying $10.32 more than Scaler's $62.92 for the same data, in exchange for a lower monthly fee at small sizes.

Supabase's rate is a third of Neon's and a sixth of Turso Developer's, but it applies to provisioned disk on one instance. The 8 GB included covers development; 100 GB of provisioned disk is a resize, not a line item that grows on its own.

Neon's $35.00 is the most honest number in the table because nothing is bundled. The storage charge is metered per GB-month, and a second charge sits beside it: history storage at $0.20 per GB-month, billed on the write-ahead log retained inside your restore window. A busy 100 GB database with a 7-day window pays more than $35.

## How we measured it

Rates and included allowances come from the three pricing pages, all opened on 2026-10-25. Nothing was provisioned and no benchmark was run. Every total multiplies 100 GB minus the published allowance by the published rate, adds the published plan fee where one exists, and is marked calculated where it is arithmetic rather than a quoted figure. Turso's plan fees are the annual-billing rates its page prints, $4.99 and $24.92, used here as the monthly fee.

## What changes in practice

The first trap is that storage is provisioned on one vendor and metered on another. Supabase bills the disk size you choose and keeps billing it whether or not it is full; Neon bills dollars per gigabyte-month actually used. A database that grows from 90 GB to 100 GB costs $1.25 more a month on Supabase after a resize and $3.50 more on Neon automatically.

The second is that Turso's plan structure inverts with size. Developer at $4.99 is the cheapest entry, but its 9 GB allowance and $0.75 rate make it the worst deal at 100 GB. Scaler at $24.92 pays for itself at this volume and has a lower overage rate for row reads as well.

The third is what is not in the table. Neon's history storage, Supabase's egress at $0.09 per GB beyond 250 GB, and Turso's sync allowance all scale with usage rather than with the database size, and each can exceed the storage line at 100 GB for a read-heavy product.

The fourth is what a resize costs in process rather than money. Supabase bills provisioned disk, so reaching 100 GB is a resize on one instance; Neon and Turso meter usage and need no resize at all, and their higher per-gigabyte rates are what that flexibility costs.

## Bottom line

At 100 GB, Supabase Pro is the cheapest storage by a wide margin at $11.50 for the line and $36.50 all in. If the workload is SQLite-shaped and reads dominate, Turso Scaler at $62.92 is the sane Turso tier and Developer is the wrong one at this size. Neon's $35.00 is fair for metered storage and becomes expensive only once history storage and always-on compute sit beside it.

**Sources:**
- Supabase pricing: https://supabase.com/pricing
- Neon pricing: https://neon.com/pricing
- Turso pricing: https://turso.tech/pricing

---

# What 1 TB of Postgres Storage Costs per Month: Supabase vs Neon vs Turso

URL: https://actualdatabase.com/postgres-storage-cost-per-tb-supabase-neon-turso/
Verified: 2026-10-25

**Quick answer:** 1 TB of database storage costs about $127/month on Supabase Pro at $0.125 per GB above the included 8 GB, about $358/month on Neon at $0.35 per GB-month, and about $761/month on Turso Developer at $0.75 per GB above its included 9 GB.

- Supabase Pro includes 8 GB of disk per project and charges $0.125 per GB beyond it (checked 2026-10-25).
- Neon charges $0.35 per GB-month of database storage and $0.20 per GB-month of history storage on paid plans (checked 2026-10-25).
- Turso Developer includes 9 GB and charges $0.75 per GB beyond it (checked 2026-10-25).
- 1 TB (1,024 GB) of storage works out to $127.00 on Supabase Pro, $358.40 on Neon and $761.25 on Turso Developer (calculated 2026-10-25).
- Supabase Pro is $25/month and includes $10 of compute credit, enough for one Micro instance (checked 2026-10-25).

Storage pricing is the quietest line on a database bill and the one that compounds. A product that grows from 10 GB to 1 TB does not notice for a year, then notices every month.

Here is what that terabyte costs on the three managed databases most side projects pick between, computed from the rates published on 2026-10-25.

## The storage line at 1 TB

| Provider | Included | Overage rate | 1 TB costs |
|---|---|---|---|
| Supabase Pro | 8 GB disk per project | $0.125/GB | $127.00 |
| Neon (Launch) | None — billed from zero | $0.35/GB-month | $358.40 |
| Turso Developer | 9 GB | $0.75/GB | $761.25 |

The arithmetic is 1,024 GB minus whatever the plan includes, times the rate. Supabase also bills file storage separately at $0.0213/GB beyond 100 GB, and Neon bills history storage at $0.20/GB-month on top of the database storage line.

[[ad]]

## What each vendor is actually selling

Supabase sells a provisioned disk attached to one Postgres instance. You choose the size, you pay for the size, and the number is predictable: the storage line moves in steps when you resize.

Neon sells storage in GB-months and separates the data from its history. The database itself is $0.35 per GB-month; the write-ahead log that makes instant restore work is $0.20 per GB-month, billed on whatever history window you configure. A team that keeps a 7-day restore window pays for seven days of writes on top of the data.

Turso stores a SQLite database per tenant rather than one big Postgres cluster. That architecture makes small databases nearly free to host and makes the per-GB rate the highest of the three, because each database carries its own overhead.

## What 1 TB does to the rest of the bill

Storage is rarely what you notice first. On Neon, a database kept always-on at the minimum 0.25 CU burns about 182.5 CU-hours a month — roughly $19.35 at $0.106 per CU-hour — before any storage. On Supabase, the $25 Pro plan includes $10 of compute credit, and the next instance size up lists at $60/month, so compute can quadruple a bill while storage is still in single digits.

The crossover that matters: below about 50 GB, compute and fixed plan fees decide the bill. Above that, storage starts to dominate on Turso and Neon, and stays a rounding error on Supabase until the terabyte range.

## How we measured it

We took the per-GB storage rates and included allowances from each vendor's pricing page on 2026-10-25 and multiplied by 1,024 GB. Nothing was provisioned and no benchmark was run, so these are list prices before negotiated discounts or annual commitments. Where a vendor bills a second storage-like line — Supabase file storage, Neon history storage — it is named in the table rather than folded into the total.

## Bottom line

For a conventional Postgres application that will hold a terabyte, Supabase's $0.125 per GB is the cheapest of the three by a factor of nearly three. Neon's higher rate buys branching, scale-to-zero and per-branch storage that only counts changed data. Turso's price only makes sense when the database count is high and each one is small — at which point the included allowances, not the overage rate, are what you are buying.

**Sources:**
- Supabase pricing: https://supabase.com/pricing
- Neon pricing: https://neon.com/pricing
- Turso pricing: https://turso.tech/pricing

---

# Railway Hobby vs Pro in 2026: $5, $20 and no per-seat charge

URL: https://actualdatabase.com/railway-hobby-vs-pro-pricing/
Verified: 2026-10-25

**Quick answer:** Railway Hobby is a $5 monthly minimum that includes $5 of usage credits, and Pro is a $20 monthly minimum that includes $20 of credits plus unlimited workspace seats. Neither is a per-seat price: the difference is in limits, with 48 vCPU and 5 GB of storage per service on Hobby against 1,000 vCPU and 1 TB on Pro.

- Railway Hobby is a $5 minimum usage charge that includes $5 of monthly usage credits, so a quiet month costs exactly $5 (checked 2026-10-25).
- Railway Pro is a $20 minimum usage charge that includes $20 of monthly usage credits and unlimited workspace seats, with no per-seat fee (checked 2026-10-25).
- Hobby allows up to 48 vCPU and 48 GB of RAM per service, while Pro allows up to 1,000 vCPU and 1 TB per service (checked 2026-10-25).
- Volume storage runs to 5 GB on Hobby and 1 TB on Pro, and log history goes from 7 days to 30 (checked 2026-10-25).
- Usage above the minimum is billed at $0.00000386 per GB-second of memory and $0.00000772 per vCPU-second of CPU, about $10.01 and $20.01 per month respectively (calculated 2026-10-25).
- A 5-person team on Pro pays the same $20 floor as one developer, an effective $4 per seat at that size (calculated 2026-10-25).

Railway sells two plans to individuals and small teams, and the surprising part in 2026 is that neither is priced per head. Hobby is a $5 monthly minimum that comes with $5 of usage credits; Pro is a $20 monthly minimum that comes with $20 of credits and unlimited workspace seats. The money is therefore almost identical for a one-person project, and the real choice is about limits: how large a single service can get, how much disk it can hold, and how far back the logs reach. Both figures below were read from Railway's pricing page on 2026-10-25.

## Hobby versus Pro

| Item | Hobby at $5 | Pro at $20 |
|---|---|---|
| Minimum usage | $5 per month | $20 per month |
| Included credits | $5 of usage | $20 of usage |
| Seats | single developer workspace | unlimited |
| CPU per service | up to 48 vCPU | up to 1,000 vCPU |
| RAM per service | up to 48 GB | up to 1 TB |
| Replicas | 5, at 8 vCPU / 8 GB | 42, at 24 vCPU / 24 GB |
| Volume storage | up to 5 GB | up to 1 TB |
| Ephemeral disk | 100 GB | 100 GB |
| Projects | 50 | 100 |
| Services per project | 50 | 100 |
| Log history | 7 days | 30 days |
| Availability target | 99.9% | 99.99% |

The free tier is a trial, not a plan: 30 days with $5 of credits, then $1 a month for 1 project, 3 services and 1 vCPU / 0.5 GB per service. Everything that looks like a real deployment starts at Hobby.

[[ad]]

## What the $20 actually means

There is no per-seat price to look up. Pro's line reads "unlimited workspace seats included", which means the bill is dominated by resource usage rather than headcount, and a five-developer workspace pays the same $20 floor as a solo developer. That is $4 per seat at five people and $20 per seat at one, so the plan gets cheaper per person precisely as the team grows.

Because the minimum is a credit rather than a fee, the two plans converge whenever usage is real. A service that burns $21.80 of resources costs $21.80 on Hobby and $21.80 on Pro — the $5 versus $20 difference only exists for accounts whose usage sits below $20, which is to say accounts that are barely running anything. The upgrade argument is limits and support, not price.

The limits are genuinely wide on both sides. Forty-eight vCPU on a Hobby service is beyond what most production apps will ever request, so the practical Pro triggers are storage past 5 GB and log history past 7 days. Those two lines are where a growing project actually feels the plan.

## How we measured it

The plan table, the resources-and-limits table and the usage-rate table were fetched on 2026-10-25 from railway.com/pricing, with the plans reference page in Railway's docs used to confirm that a base subscription price is credited against usage. No service was deployed and no invoice was generated; per-second rates were multiplied by 2,592,000 seconds, the length of a 30-day month, and those conversions are labelled as calculated.

## What changes in practice

The billing switch at the minimum is the trap. On Hobby, once usage passes $5 you are charged the total usage, not the usage plus $5, so the mental model of "plan fee plus metered usage" is wrong by exactly the plan fee. A month at $21.80 bills $21.80, and a month at $8 bills $8 — the floor only applies when usage is lower than the floor.

Memory is the line that surprises people, because it bills whether or not the service is busy. At $0.00000386 per GB-second, holding 1 GB costs about $10.01 a month even at zero traffic, so a forgotten staging environment is a real invoice. Railway does not charge for a service that is not running, which makes deleting idle environments the single largest cost control available.

Storage is the other Pro trigger. Five gigabytes of volume space is plenty for a small database and nowhere near enough for user uploads or a media cache; the Hobby ceiling arrives before the compute ceiling does. And since object storage bills at $0.015 per GB-month with free egress, moving files out of a volume and into a bucket is usually cheaper than upgrading.

## Bottom line

Pick Hobby at $5 for anything with one developer and under 5 GB of data, and move to Pro at $20 when you need a second person, 30-day logs or more than 5 GB of volume. Because usage above the minimum is charged identically on both plans, the $15 difference is really a limits fee, and it buys 195 times the storage allowance (calculated 2026-10-25).

**Sources:**
- Railway pricing: https://railway.com/pricing
- Railway docs: pricing plans: https://docs.railway.com/reference/pricing/plans

---

# Railway usage billing in 2026: $10.01 per GB-month and $20.01 per vCPU-month

URL: https://actualdatabase.com/railway-usage-based-billing-explained/
Verified: 2026-10-25

**Quick answer:** Railway meters memory at $0.00000386 per GB-second, CPU at $0.00000772 per vCPU-second, volumes at $0.00000006 per GB-second and egress at $0.05 per GB. Over a 30-day month that is about $10.01 per GB of RAM, $20.01 per vCPU and $0.16 per GB of disk, so a service with 1 GB, half a vCPU, 20 GB of egress and 5 GB of disk bills about $21.80 (calculated 2026-10-25).

- Railway bills memory at $0.00000386 per GB-second, which is about $10.01 per GB for a 30-day month (calculated 2026-10-25).
- Railway bills CPU at $0.00000772 per vCPU-second, about $20.01 per vCPU for a 30-day month (calculated 2026-10-25).
- Volume storage bills at $0.00000006 per GB-second, about $0.16 per GB per 30-day month, and egress from services costs $0.05 per GB (calculated 2026-10-25).
- A worked example of 1 GB of RAM, 0.5 vCPU, 20 GB of egress and 5 GB of volume bills about $21.80 a month, of which $20.02 is memory and CPU (calculated 2026-10-25).
- Hobby includes $5 of usage credit and Pro includes $20, and once usage passes the minimum the full usage is billed rather than the difference (checked 2026-10-25).
- Object storage is the exception at $0.015 per GB-month with free egress, against $0.05 per GB of egress from a service (checked 2026-10-25).

Railway does not sell servers, it sells seconds. Memory is metered per GB-second, CPU per vCPU-second, storage per GB-second and traffic per gigabyte, which means the only way to predict a bill is to convert those rates into monthly figures and then guess how busy the app will be. The conversions are simple: about $10.01 per GB of RAM, $20.01 per vCPU, $0.16 per GB of disk and $0.05 per GB of egress. A service holding 1 GB, using half a vCPU, sending 20 GB and keeping 5 GB of disk lands at about $21.80 for the month.

## The four meters

| Meter | Published rate | Cost per month at 30 days |
|---|---|---|
| Memory | $0.00000386 per GB/sec | $10.01 per GB |
| CPU | $0.00000772 per vCPU/sec | $20.01 per vCPU |
| Volume storage | $0.00000006 per GB/sec | $0.16 per GB |
| Egress from a service | $0.05 per GB | $51.20 per TB |
| Object storage | $0.015 per GB-month | egress free |
| Sandbox memory and CPU | $0.00001929 per unit/sec | about $50 per GB or vCPU |

The monthly column assumes 2,592,000 seconds, which is 30 days exactly. A 31-day month is 2,678,400 seconds and produces a bill about 3.3 percent higher for the same usage, so annual budgeting should use the longer number.

[[ad]]

## Working the $21.80 example

Take a Node service that holds 1 GB of RAM, averages half a vCPU, sends 20 GB to the internet and keeps 5 GB of volume. Memory is 1 times $10.01, CPU is 0.5 times $20.01, egress is 20 times $0.05 and storage is 5 times $0.16. That is $10.01 plus $10.01 plus $1.00 plus $0.78, or $21.80.

Notice which line dominates: memory and CPU together are $20.02 of the $21.80, and egress is a dollar. Railway is a compute-priced platform, and bandwidth is almost decorative next to it. A team that optimises image sizes before it right-sizes memory is optimising the wrong line by a factor of twenty.

The same arithmetic explains why the plan minimum rarely matters for real apps. $5 of credit on Hobby covers 512 MB of memory for a month and nothing else, so anything with a second service or a database crosses the minimum immediately and is billed on usage instead of on the plan.

## How we measured it

Railway's pricing page was fetched on 2026-10-25 and every per-second rate was multiplied by 2,592,000 to produce the monthly figures shown above. The plans reference page in Railway's docs was used to confirm that the base subscription is a credit against usage rather than a fee on top of it. No service was deployed, no sandbox was run and no invoice was inspected, so the worked example is arithmetic on published rates plus stated assumptions about CPU and traffic.

## What changes in practice

The billing switch above the minimum is the single most misunderstood part. Railway charges total usage once usage exceeds the plan minimum, so the bill is not plan fee plus usage. A month at $21.80 of usage bills $21.80 on Hobby, which means the $5 minimum is a floor and never a fixed cost — and it also means upgrading to Pro at $20 does not reduce a heavy bill, it only raises the floor.

Memory bills whether or not traffic arrives, and that is the cost that grows silently. Three forgotten preview environments at 1 GB each are $30.03 a month, more than the plan itself. Railway does not charge for a service that is stopped, so deleting environments is the cheapest optimisation available, ahead of any code change.

Sandboxes have their own, much higher, rates: $0.00001929 per GB-second is roughly $50 per GB-month, five times a service. They bill only while running and an idle timeout destroys them, but a sandbox waiting on a model for an hour still costs real money, and it draws from the same included credit as everything else.

## Bottom line

Budget $10.01 per GB of RAM and $20.01 per vCPU per month, add $0.05 per GB of egress, and expect the plan minimum to be irrelevant once more than one service exists. A realistic two-service app with 1 GB total and half as much CPU averages about $25 to $30 a month, which is still below the $45 a month a comparable managed instance costs elsewhere (calculated 2026-10-25).

**Sources:**
- Railway pricing: https://railway.com/pricing
- Railway docs: pricing plans: https://docs.railway.com/reference/pricing/plans

---

# Render disk and bandwidth in 2026: $0.25 per GB of disk, $0.15 per GB of traffic

URL: https://actualdatabase.com/render-disk-and-bandwidth-cost/
Verified: 2026-10-25

**Quick answer:** Render charges $0.25 per GB per month for a persistent disk and $0.15 per GB for bandwidth beyond the workspace allowance, which is 5 GB on Hobby, 25 GB on Pro and 1 TB on Scale. A 10 GB disk is $2.50 a month and 100 GB of traffic on the $0 Hobby plan is $14.25 a month in overage (calculated 2026-10-25).

- A persistent disk on Render costs $0.25 per GB per month on every plan, so 10 GB is $2.50 and 100 GB is $25.00 (checked 2026-10-25).
- Bandwidth beyond the included allowance costs $0.15 per GB, and the allowance is 5 GB on Hobby, 25 GB on Pro and 1 TB on Scale (checked 2026-10-25).
- 100 GB of monthly traffic costs $14.25 on the $0 Hobby plan, which includes 5 GB and bills the remaining 95 GB at $0.15 (calculated 2026-10-25).
- The same 100 GB costs $36.25 on Pro once the $25 workspace fee is included, because Pro only adds 20 GB of included bandwidth worth $3.00 (calculated 2026-10-25).
- Storage is priced separately for managed databases: Postgres storage is $0.30 per GB per month with the first 1 GB free, and Key Value starts at Free 25 MB (checked 2026-10-25).
- Free instances cannot attach a persistent disk at all, so durable state starts at the $7 compute plan plus $0.25 per GB (checked 2026-10-25).

Render's pricing page splits storage and traffic into two cheap-looking unit prices that behave very differently over a year. A persistent disk is $0.25 per GB per month and it bills while it exists, whether or not anything reads it. Bandwidth is $0.15 per GB and it bills only once a workspace passes its included allowance of 5 GB on Hobby, 25 GB on Pro or 1 TB on Scale. The first is rent, the second is postage, and confusing them is how a $7 service turns into a $30 invoice.

## What a gigabyte costs

| Item | Rate | Concrete example |
|---|---|---|
| Persistent disk | $0.25 per GB/month | 10 GB = $2.50 |
| Disk, 100 GB | $0.25 per GB/month | $25.00 |
| Hobby bandwidth | 5 GB included | then $0.15 per GB |
| Pro bandwidth | 25 GB included | then $0.15 per GB |
| Scale bandwidth | 1 TB included | then $0.15 per GB |
| 100 GB month on Hobby | 95 GB over | $14.25 |
| Postgres storage | $0.30 per GB/month | first 1 GB free |
| Custom domains, Hobby | 2 included | then $0.25 each |

Everything on this page is prorated per second, so a disk created on the 20th costs about a third of its monthly rate for that month.

[[ad]]

## What the two meters actually mean

A disk makes a service stateful, and that is a different product from a stateless web process. Render's documentation is explicit that without a disk, every file written to the local filesystem is lost on the next deploy or restart, and that only data written under the disk's mounting path survives. Disks are encrypted at rest and take automatic daily snapshots, which is part of what the $0.25 buys. The other part is that they are unavailable on Free instances, so the first gigabyte of durable state also buys the $7 compute plan.

Bandwidth works in the opposite direction: it charges for consumption, not for standing still. The Hobby workspace includes 5 GB and then bills each gigabyte at $0.15. That allowance is shared across every service in the workspace, so a static site and an API and a preview environment all draw from the same 5 GB. Static deploys are free; the traffic they serve is not.

The interaction between the two explains most surprising bills. Server-rendered pages with large images push bandwidth without touching disk. Uploads and generated PDFs push disk without touching bandwidth. A service doing both pays on two meters, and the one that grows quietly is disk, because it never shrinks when traffic drops.

## How we measured it

Render's pricing page was fetched on 2026-10-25 for the per-unit rates, the included allowances and the workspace prices, and Render's persistent disks documentation was fetched the same day for the behaviour of disks and their limits. Both totals in this piece are arithmetic on those published rates, assuming 100 GB of traffic and a 100 GB disk. No service was deployed and no invoice was inspected.

## What changes in practice

Upgrading to Pro to save bandwidth does not work. Pro costs $25 a month and raises the included allowance from 5 GB to 25 GB, which is worth at most $3.00 of overage, so the $22 difference is paid for the other features: unlimited seats and services, 15 custom domains, 1,000 build minutes and 14-day logs. A 100 GB month is $14.25 on Hobby and $36.25 on Pro, which is the arithmetic version of the same point (calculated 2026-10-25).

Custom domains behave like disks in miniature: they are included in ones and twos and then billed at $0.25 each per month. Hobby includes 2, Pro 15, Scale 25, and six domains on Hobby therefore adds $1.00. It is small money, but it belongs in the same column as disk rather than the same column as bandwidth, because it recurs whether or not anyone visits.

The trap worth avoiding is using a disk for data that belongs in a database. Postgres storage is $0.30 per GB per month with the first gigabyte free, and it brings backups and a query interface; a 100 GB disk costs $25 and gives you a filesystem you have to back up yourself. For relational data the managed service is cheaper per gigabyte of usable durability, and the disk is for the things a database is bad at.

## Bottom line

Budget $0.25 per GB of disk per month and $0.15 per GB of traffic past the allowance, and expect a real small app with 10 GB of uploads and 50 GB of traffic to add about $9.25 to a $7 compute bill, or roughly $16.25 all in (calculated 2026-10-25).

**Sources:**
- Render pricing: https://render.com/pricing
- Render docs: persistent disks: https://render.com/docs/disks

---

# Render Free vs Starter in 2026: a $0 service that sleeps, or $7 that does not

URL: https://actualdatabase.com/render-free-vs-starter-pricing/
Verified: 2026-10-25

**Quick answer:** Render's Free web service costs $0 but spins down after 15 minutes without inbound traffic and takes about 1 minute to wake, while the Starter instance at $7 a month stays up and adds a 0.5 CPU / 512 MB allocation. The $7 Starter tier is the first paid step in Render's ladder, and the jump to the next size is $25 for 1 CPU and 2 GB.

- Render's Free web service costs $0 and spins down after 15 minutes without inbound traffic, waking in about 1 minute on the next request (checked 2026-10-25).
- A Render workspace gets 750 Free instance hours per calendar month, and spun-down services do not consume them (checked 2026-10-25).
- The Starter instance is 0.5 CPU and 512 MB for $7 a month and runs continuously, with no idle spin-down (checked 2026-10-25).
- Free includes 5 GB of workspace bandwidth and 500 build minutes a month, with extra bandwidth at $0.15 per GB and extra builds at $5 per 1,000 minutes (checked 2026-10-25).
- Free web services cannot attach a persistent disk, scale past 1 instance, or use SSH, one-off jobs or edge caching (checked 2026-10-25).
- Free Postgres gives 1 GB of storage and expires 30 days after creation, and the paid database ladder starts at $6 a month (checked 2026-10-25).

Render's cheapest two options are not the same product at different prices. The Free web service is $0 and disappears after 15 minutes of quiet, while the Starter instance is $7 a month and behaves like a server. Between them sit the things that decide most small projects: cold starts, a persistent disk, SSH access and a working database. The figures below come from Render's pricing page and its free-tier documentation, both read on 2026-10-25.

## Free versus Starter, line by line

| Item | Free | Starter at $7 a month |
|---|---|---|
| Price | $0 | $7, prorated to the second |
| Allocation | 512 MB, less than 1 CPU | 0.5 CPU, 512 MB |
| Idle behaviour | spins down after 15 minutes | always running |
| Cold start | about 1 minute | none |
| Instance hours | 750 per workspace per month | unlimited, billed hourly |
| Bandwidth included | 5 GB per workspace | 5 GB on the Hobby workspace |
| Build minutes | 500 per workspace | 500 on the Hobby workspace |
| Persistent disk | not available | $0.25 per GB per month |
| Instances | 1, cannot scale | manually scalable |
| SSH, shell, one-off jobs | not available | available |
| Custom domains | 2 on the Hobby workspace | 2 on the Hobby workspace |

Both rows are governed by the Hobby workspace, which is $0 a month plus compute. Render waives the workspace fee in a month with no services and no activity, so a single Starter service on Hobby costs exactly $7 plus whatever bandwidth exceeds 5 GB.

[[ad]]

## What the sleep timer actually means

Fifteen minutes is shorter than most people assume. A service that gets one visitor every 20 minutes is asleep every time that visitor arrives, which means every page view in a low-traffic app pays roughly a minute of cold start. WebSocket messages count as inbound traffic, so a chat app with a connected client stays awake; a documentation site with real readers does not.

The 750-hour allowance is more generous than it looks and less generous than it sounds. Two always-on Free services would need 1,440 hours in a 30-day month, so a workspace can keep exactly one Free service awake around the clock and still run out — the hours are meant for services that actually sleep. When they run out, every Free web service in the workspace is suspended until the next month, and the balance does not roll over.

The database is the part that ends most Free experiments. Free Postgres is 1 GB and expires 30 days after creation, with a 14-day grace period and no backups, so an app that lives on Free for a quarter has to move its data twice: once off the expired database and once onto something with a bill attached.

## How we measured it

Three Render pages were fetched on 2026-10-25: the pricing page for the compute ladder, the free-tier docs for spin-down behaviour, instance hours and blocked features, and the compute-plan docs to confirm that Starter is the current name for the 0.5 CPU / 512 MB instance. No service was deployed and no cold start was timed — the one-minute figure is Render's own, and every derived number, such as 750 hours being 31.25 days, is marked as calculated.

## What changes in practice

Paying $7 does not buy more RAM, because Starter and Free are both 512 MB. It buys continuity, a disk and a shell. That distinction matters when people upgrade expecting a faster app and instead get a warmer one: the win is that the first request after lunch no longer takes a minute, not that the app computes faster.

The upgrade also unlocks the operational features that make a service debuggable. SSH and one-off jobs mean you can run a migration against the live environment, and a persistent disk at $0.25 per GB per month means uploaded files survive a deploy. Neither is available on Free at any price.

Watch the bandwidth line, because it is shared. The Hobby workspace includes 5 GB a month across all services and static sites, then charges $0.15 per GB, so a Free service serving 20 GB of images costs $2.25 in overage while the compute stays at $0. That is the cheapest part of Render and the easiest to forget.

## Bottom line

Start on Free if the app can tolerate a 1-minute cold start and stores nothing on local disk, and move to the $7 Starter instance the first time a real user complains about the wait. Skipping to $25 for 1 CPU and 2 GB only makes sense when the app is actually short of memory, which for most side projects it is not (calculated 2026-10-25).

**Sources:**
- Render pricing: https://render.com/pricing
- Render docs: free tier limits: https://render.com/docs/free
- Render docs: compute plan types: https://render.com/docs/compute-plans

---

# Render vs Railway in 2026: a 512 MB app costs $7.75 against $7.50

URL: https://actualdatabase.com/render-vs-railway-for-a-small-app/
Verified: 2026-10-25

**Quick answer:** A single 512 MB web service with 10 GB of monthly traffic costs about $7.75 on Render and about $7.50 on Railway's Hobby plan. Render charges a flat $7 for 0.5 CPU and 512 MB, while Railway bills memory at $10.01 per GB-month, CPU at $20.01 per vCPU-month and egress at $0.05 per GB against a $5 minimum (calculated 2026-10-25).

- The same 512 MB web service with 10 GB of monthly egress costs about $7.75 on Render and about $7.50 on Railway's Hobby plan (calculated 2026-10-25).
- Render charges a flat $7 a month for 0.5 CPU and 512 MB, so the bill does not move when the service is idle (checked 2026-10-25).
- Railway bills 512 MB of memory at about $5.00 a month and egress at $0.05 per GB, against $0.15 per GB over Render's 5 GB allowance (calculated 2026-10-25).
- Railway's $5 minimum means a nearly idle 512 MB service still costs $5 a month, only $2 below Render's flat $7 (checked 2026-10-25).
- Railway overtakes Render's $7.75 at about 0.11 vCPU of average use with 10 GB of egress, so the cheaper platform depends on measured CPU (calculated 2026-10-25).
- At 100 GB of egress the gap widens to about $12.00 on Railway against $21.25 on Render, because bandwidth overage is three times cheaper (calculated 2026-10-25).

One web service, 512 MB of RAM, modest traffic: that is the shape of most side projects, and pricing it on both platforms produces totals within 25 cents of each other. Render sells a fixed 0.5 CPU / 512 MB instance for $7 a month. Railway sells seconds of memory and CPU against a $5 minimum, which for 512 MB works out to about $5.00 of memory plus whatever CPU the app actually uses. Priced at 10 GB of monthly egress and an assumed 0.1 vCPU average, Render comes to $7.75 and Railway to $7.50.

## The same app, both invoices

| Line item | Render | Railway |
|---|---|---|
| Workspace or minimum | $0 Hobby workspace | $5 minimum usage |
| 512 MB of RAM | $7.00 flat, 0.5 CPU included | $5.00 at $10.01 per GB-month |
| CPU | included in the $7 | $2.00 at an assumed 0.1 vCPU |
| 10 GB of egress | $0.75, after 5 GB free | $0.50 at $0.05 per GB |
| Monthly total | $7.75 | $7.50 |
| Idle cost | $7.00 | $5.00 |
| Next size up | $25 for 1 CPU / 2 GB | no plan change, metered |

The 0.1 vCPU figure is an assumption, not a published number, because Railway bills measured CPU and Render does not measure it at all. That single assumption is what decides which platform is cheaper, and it is worth stating plainly rather than hiding inside a total.

[[ad]]

## What the 25-cent gap actually means

Render's price is a ceiling and Railway's is a forecast. Paying $7 on Render buys 0.5 CPU whether the app uses it or not, which is comfortable for a build step or a burst of image resizing and wasteful for a service that answers four requests an hour. Railway charges for the CPU that actually ran, so a nearly idle app pays close to the $5 memory floor — and a busy one pays more than Render would have charged.

The crossover is easy to locate. Railway's bill is $5.00 of memory plus $20.01 per vCPU-month plus egress, so against Render's $7.75 the CPU budget is $2.25, or about 0.11 vCPU averaged across the month (calculated 2026-10-25). Below that, Railway wins; above it, Render's flat rate is the better deal.

Bandwidth is the asymmetric line. Render includes 5 GB and charges $0.15 per GB after it, while Railway charges $0.05 per GB from the first byte but includes $5 of credit that absorbs 100 GB. At 100 GB of monthly egress the totals are about $12.00 on Railway and $21.25 on Render, which is a 1.8-times gap created entirely by the overage rate (calculated 2026-10-25).

## How we measured it

Both pricing pages were fetched on 2026-10-25 and the metered rates were converted to monthly figures by multiplying per-second prices by 2,592,000 seconds. No service was deployed on either platform and no invoice was compared, so the totals here are arithmetic on published rates plus one clearly labelled CPU assumption. Render's flat price and Railway's rates both come straight from the vendors' own tables.

## What changes in practice

Render's minimum is per workspace but its compute is per service, so the second 512 MB service adds another $7 rather than fitting inside the first invoice. Railway's minimum is per workspace in a different sense: two 512 MB services bill $10.01 of memory, which already exceeds the $5 floor, so the floor stops protecting you as soon as there are two services.

Storage is where the two diverge quietly. Render's persistent disks cost $0.25 per GB per month, while Railway's volumes bill at $0.00000006 per GB-second, about $0.16 per GB per month. A 20 GB volume is $5.00 on Render and about $3.12 on Railway, which is larger than the compute difference at these sizes (calculated 2026-10-25).

Neither platform charges for a build that nobody triggers, but both charge for the app sitting there. That makes deleting staging environments the highest-leverage cost control on either one, and on Railway it is the only thing standing between a forgotten service and a $5 memory charge every month.

## Bottom line

Choose Render at $7.75 if you want a number that does not move and a CPU allowance you never have to watch, or Railway at $7.50 if traffic is heavy or the app is idle most of the day. If the app genuinely idles, the honest budget option is neither: a $4 DigitalOcean Droplet beats both, at the cost of running the server yourself (calculated 2026-10-25).

**Sources:**
- Render pricing: https://render.com/pricing
- Railway pricing: https://railway.com/pricing

---

# Resend Pro vs Scale: the $90 tier never gets cheaper per email

URL: https://actualdatabase.com/resend-pro-vs-scale-break-even/
Verified: 2026-10-25

**Quick answer:** Never on price: Pro is $20/month for 50,000 emails and Scale is $90/month for 100,000, and both charge $0.90 per 1,000 emails above their allowance. Scale's extra 50,000 emails cost $70 while the same volume costs $45 in overage on Pro, so Pro is $25/month cheaper at every volume above 100,000 emails.

- Resend Pro costs $20/month for 50,000 emails and Scale costs $90/month for 100,000 emails (checked 2026-10-25).
- Both paid plans charge $0.90 per 1,000 extra emails, billed in 1,000-email buckets (checked 2026-10-25).
- At 100,000 emails a month, Pro totals $65.00 against Scale's $90.00 (calculated 2026-10-25).
- At 200,000 emails a month, Pro totals $155.00 against Scale's $180.00, so Pro stays $25.00/month cheaper above 100,000 (calculated 2026-10-25).
- Scale's extra 50,000 emails cost $70.00 of plan fee but are worth $45.00 at the $0.90 overage rate, which is where the $25.00 gap comes from (calculated 2026-10-25).
- Dedicated IPs are a $30/month add-on offered to customers exceeding 3,000 emails per day on the Scale plan (checked 2026-10-25).

Resend sells four tiers, and the gap between the second and third looks like a volume discount until you do the division. Pro is $20 a month for 50,000 emails. Scale is $90 a month for 100,000. Both then charge the same $0.90 per 1,000 extra.

Doubling the allowance costs 4.5 times the price. Run the arithmetic at any volume above 100,000 and Pro plus overage comes out $25 a month cheaper than Scale, which makes Scale a features purchase rather than a volume one.

## Pro against Scale, with overage priced in

| Emails per month | Pro total | Scale total | Cheaper |
|---|---|---|---|
| 50,000 | $20.00 | $90.00 | Pro by $70.00 |
| 60,000 | $29.00 | $90.00 | Pro by $61.00 |
| 100,000 | $65.00 | $90.00 | Pro by $25.00 |
| 150,000 | $110.00 | $135.00 | Pro by $25.00 |
| 200,000 | $155.00 | $180.00 | Pro by $25.00 |

Totals are the plan fee plus $0.90 for every 1,000 emails above the allowance. Beyond 100,000 the two columns rise in parallel, one dollar apart per 1,000 emails, so the $25 gap never closes.

[[ad]]

## What the numbers actually mean

Scale's premium buys 50,000 additional emails for $70. At the rate Resend itself charges for extra emails, those 50,000 are worth $45. The remaining $25 is the price of everything else in the tier.

That is the whole break-even analysis, and it has no crossover point. A sender at 500,000 emails a month pays $20 + 450 x $0.90 = $425 on Pro and $90 + 400 x $0.90 = $450 on Scale. A sender at 2 million pays $1,775 against $1,800. A sender at 10 million emails pays $8,975 on Pro against $9,000 on Scale, still $25 apart. The gap is structural: it comes from the ratio between the two plan fees and the two allowances, not from the volume.

What does change with volume is the effective rate. Pro at 50,000 emails is $0.40 per 1,000. Pro at 100,000 is $0.65. Scale at 100,000 is $0.90, and Scale at 200,000 is $0.675. Pro's effective rate only reaches Scale's headline rate at exactly 50,000 emails of overage beyond its allowance.

## How we measured it

Plan prices, allowances and the $0.90 overage rate are quoted from Resend's pricing page on 2026-10-25. The 1,000-email billing bucket and the pay-as-you-go requirement come from the FAQ on that page. Every total in the table is plan fee plus overage at those published rates and is marked calculated; nothing was sent and no deliverability test was run.

## What changes in practice

The practical risk of staying on Pro is not price, it is the switch. Resend bills overage only when pay-as-you-go is enabled under Transactional Overages; otherwise sending stops at 50,000 emails. A product that grows through a launch week meets that ceiling on a Tuesday afternoon with no warning beyond the dashboard.

The practical risk of moving to Scale is paying for features you will not use. The tier's published advantages are administrative: 1,000 custom domains against 10, 10 webhook endpoints against 5, 500 AI credits against 100, and eligibility for the $30 a month dedicated IP add-on.

The dedicated IP is the one hard dependency. Resend offers it only to customers exceeding 3,000 emails sent per day on the Scale plan, so a sender with deliverability problems at that volume has to pay the $25 premium to buy the add-on. That is a defensible reason to be on Scale; saving money is not.

The free plan is not a substitute for either. It carries 3,000 emails a month with a 100-email daily cap, so a product sending 50,000 emails cannot serve any meaningful share of that volume without paying. At 50,000 emails Pro's effective rate is $0.40 per 1,000 and Scale's is $1.80, because Scale bills for its whole 100,000-email allowance whether or not the volume is used.

## Bottom line

For volume alone, stay on Pro and enable overage: at 100,000 emails it is $65 against $90, and the $25 gap holds at every higher volume. Move to Scale only when you need 1,000 domains, the dedicated IP add-on, or the higher webhook and AI allowances — and treat the $30 IP charge as the real cost of that decision.

**Sources:**
- Resend pricing: https://resend.com/pricing
- Resend docs: which sending feature to use: https://resend.com/docs/knowledge-base/what-sending-feature-to-use

---

# Resend vs Postmark: What Transactional Email Costs in 2026

URL: https://actualdatabase.com/resend-vs-postmark-transactional-email-pricing/
Verified: 2026-10-25

**Quick answer:** Resend is cheaper at every volume above a few thousand emails: $20/month for 50,000 emails with $0.90 per extra 1,000, against Postmark's $15/month for 10,000 emails with $1.80 per extra 1,000. At 50,000 emails a month that is $20 versus roughly $87 at Postmark's published overage rate.

- Resend Free: 3,000 emails per month, capped at 100 emails per day, $0 (checked 2026-10-25).
- Resend Pro: $20/month for 50,000 emails, then $0.90 per extra 1,000 (checked 2026-10-25).
- Postmark Basic: $15/month for 10,000 emails, then $1.80 per extra 1,000 (checked 2026-10-25).
- Postmark's free developer plan is 100 emails per month and does not expire (checked 2026-10-25).
- 50,000 emails a month costs $20 on Resend Pro and about $87 on Postmark Basic at its $1.80 overage rate (calculated 2026-10-25).
- Resend's dedicated IP add-on is $30/month; Postmark's dedicated IPs start at $50/month and require 300,000 emails per month (checked 2026-10-25).

Both Resend and Postmark sell the same thing — a reliable API that puts a password reset in an inbox instead of a spam folder — and both price it per email. The plans diverge sharply once you leave the entry tier.

Here are the published numbers as of 2026-10-25, and what the same monthly volume bills on each.

## Entry tiers, side by side

| | Resend | Postmark |
|---|---|---|
| Free plan | 3,000 emails/month, 100/day | 100 emails/month, no expiry |
| First paid tier | $20/month, 50,000 emails | $15/month, 10,000 emails |
| Overage rate | $0.90 per extra 1,000 | $1.80 per extra 1,000 |
| Next tier up | $90/month, 100,000 emails | $16.50/month, $1.30 per extra 1,000 |
| Volume ceiling | 100,000 included before overage | Listed tiers step to $1.20 per extra 1,000 |
| Dedicated IP | $30/month add-on | From $50/month, requires 300,000 emails/month |
| Extra domains | $20/month for 100 more | Included |

[[ad]]

## What the same volume actually bills

| Emails per month | Resend | Postmark (at Basic overage) |
|---|---|---|
| 100 | $0 | $0 |
| 3,000 | $0 | $15 |
| 10,000 | $20 | $15 |
| 25,000 | $20 | $42 |
| 50,000 | $20 | $87 |
| 100,000 | $90 | $177 |

The Postmark column is arithmetic on the published Basic rate: $15 plus $1.80 for every additional 1,000. Postmark's pricing slider shows the per-1,000 rate dropping to $1.30 and then $1.20 at higher volumes, so the real 100,000-email bill sits below that $177 figure. Resend's column is its published plan price, with $0.90 per 1,000 only once a plan's included volume is used up.

## How we measured it

Both pricing pages were loaded on 2026-10-25 and the numbers read out of the plan tables, not the marketing copy. Overage arithmetic is plain multiplication; no benchmark was run and no discount was assumed. Where a vendor answers a pricing question in its FAQ rather than in the table — Postmark's "no tier between 100 and 10,000", Resend's "no monthly-to-annual switch" — that answer is quoted rather than restated as a number.

## What changes in practice

The crossover is narrow and early. Postmark is cheaper below roughly 13,000 emails a month, and above that Resend's included volume wins until Postmark's volume discounts close the gap. A product sending 10,000 emails a month — password resets and receipts for a few thousand users — is paying $15 either way, so price is not the deciding factor at that size.

The expensive part of email is not the per-email rate. It is deliverability: shared IP reputation, the receiving provider's filtering, and the first week after you switch providers. Postmark's reputation is the reason people pay its overage rates, and Resend's shared IP pools are the reason its entry tier is cheap. Neither vendor publishes a deliverability number you can compare, so this is the one line in the table where the price is not the whole story.

Watch the add-ons. A dedicated IP costs $30/month on Resend and starts at $50/month on Postmark, and Postmark only sells it to accounts sending 300,000 emails a month or more. On Resend, 100 extra domains cost $20/month, which is more than the Pro plan itself.

## Bottom line

Under about 13,000 emails a month the two are within a few dollars, and Postmark's tooling is the tiebreaker. From 20,000 upward, Resend's included volume is cheaper by a factor that grows with usage — about 4x at 50,000 emails — unless Postmark's volume tier and reputation justify the difference for your sending pattern.

**Sources:**
- Resend pricing: https://resend.com/pricing
- Postmark pricing: https://postmarkapp.com/pricing

---

# What 1 TB of Cloud Egress Costs: S3 vs Backblaze B2 vs Wasabi

URL: https://actualdatabase.com/s3-vs-backblaze-b2-vs-wasabi-egress-cost/
Verified: 2026-10-25

**Quick answer:** 1 TB of egress from Amazon S3 costs about $83 after the 100 GB monthly free tier, because AWS charges $0.09 per GB for data transfer out of S3 in Europe (Ireland). Backblaze B2 charges $0 while egress stays under 3x your stored data and $0.01/GB after, and Wasabi charges nothing for egress on a $7.99 per TB per month plan.

- AWS quotes $0.09 per GB for data transfer out of S3 to the internet in Europe (Ireland) (checked 2026-10-25).
- AWS gives 100 GB per month of data transfer out to the internet free, aggregated across all AWS services and regions (checked 2026-10-25).
- 1 TB of egress from S3 therefore bills about $83: 924 billable GB x $0.09 (calculated 2026-10-25).
- Backblaze B2 costs $6.95 per TB per month for storage, with free egress up to 3x your average monthly stored data and $0.01/GB beyond it (checked 2026-10-25).
- Wasabi costs $7.99 per TB per month with no egress or API fees, a 90-day minimum storage duration, and a minimum monthly charge tied to 1 TB of active storage per account (checked 2026-10-25).

Storage is cheap and egress is not. A terabyte of object storage costs a few dollars a month on every provider; pulling that terabyte back out over the public internet costs nothing on two of them and around $83 on S3.

The three numbers below come from the vendors' own pricing pages on 2026-10-25, including the free allowances that decide the final bill.

## The egress bill for 1 TB of downloads

| Provider | Storage price | Free egress | 1 TB of egress costs |
|---|---|---|---|
| Amazon S3 | Per-region, billed per GB-month | 100 GB/month, aggregated across AWS | About $83 (924 GB x $0.09) |
| Backblaze B2 | $6.95 per TB/month | Up to 3x average monthly stored data | $0, or $0.01/GB beyond 3x |
| Wasabi | $7.99 per TB/month | All of it | $0 |

[[ad]]

## Where the S3 number comes from

AWS's S3 pricing page states the rate in a worked example: data transfer out from Amazon S3 in Europe (Ireland) to the internet is $0.09 per GB. The same page states that AWS customers get 100 GB of data transfer out to the internet free each month, aggregated across all AWS services and regions.

So a terabyte of egress — 1,024 GB — is 924 billable gigabytes, and 924 x $0.09 = $83.16. Move the workload to a higher-volume tier and the per-GB rate falls; move it to a cheaper region and the rate changes. The arithmetic, not the rate, is the part people get wrong.

## How B2 and Wasabi avoid the charge

Backblaze B2 charges $6.95 per TB per month for storage and makes egress free up to three times your average monthly stored data. Store 1 TB and you can serve 3 TB of downloads a month for nothing; go past it and the overage is $0.01 per GB, so 4 TB of egress on 1 TB of storage costs about $10. B2 also waives egress entirely when the download goes through a listed CDN or compute partner — Fastly, Cloudflare, bunny.net, CacheFly, CoreWeave, Equinix Metal, Vultr and phoenixNAP.

Wasabi takes a different trade. It charges $7.99 per TB per month, includes egress and API calls at no extra cost, requires a 90-day minimum storage duration, and ties the minimum monthly charge to 1 TB of active storage per account. You pay more per stored byte than on B2 and get an unlimited download allowance for it.

## How we measured it

We read the egress rules and rates straight from the three pricing pages on 2026-10-25 and did the arithmetic in the table above by hand. No transfer benchmark was run, so these are bills, not speeds. AWS renders most individual prices dynamically on that page, which is why the S3 rate quoted here is the one printed in AWS's own worked example rather than a region table we could scrape.

## What changes in practice

Egress is a business-model tax, not an infrastructure cost. A media site, a public dataset mirror, or an API that returns large payloads pays it every month; an application that stores backups and rarely reads them pays almost nothing anywhere. That is why the same 1 TB costs $83 on S3 and $0 on Wasabi — the providers are pricing two different customer shapes.

Two clauses decide real bills more than the headline rate. B2's 3x allowance is relative to average monthly storage, so a bucket that shrinks in the middle of the month also shrinks its free egress. Wasabi's 90-day minimum means an export pipeline that writes and deletes daily still bills for three months of every object.

## Bottom line

For read-heavy public data, Wasabi's flat $7.99 per TB with free egress is the simplest number to predict. For storage-heavy, read-light workloads, B2's $6.95 per TB plus a 3x egress allowance is cheaper. S3 remains the option you pick for the ecosystem, and it is the only one of the three where a popular download can turn into a bill with four digits.

**Sources:**
- Amazon S3 pricing: https://aws.amazon.com/s3/pricing/
- Backblaze B2 pricing: https://www.backblaze.com/cloud-storage/pricing
- Wasabi pricing: https://wasabi.com/pricing

---

# Stripe Card Processing Fees in 2026: 2.9% + 30¢, Line by Line

URL: https://actualdatabase.com/stripe-card-processing-fees-2026/
Verified: 2026-10-25

**Quick answer:** Stripe charges 2.9% + 30¢ per successful domestic card charge in the United States, which is $0.59 on a $10 charge and $3.20 on a $100 charge. International cards add 1.5%, currency conversion adds 1%, and a dispute costs $15.00.

- Stripe charges 2.9% + 30¢ per successful domestic card charge in the United States (checked 2026-10-25).
- A $10 charge costs $0.59 and nets $9.41; a $100 charge costs $3.20 and nets $96.80 (calculated 2026-10-25).
- The fixed 30¢ is 50.8% of the fee on a $10 charge and 9.4% of the fee on a $100 charge (calculated 2026-10-25).
- International cards add 1.5% and currency conversion adds 1%, taking a $100 charge to $5.70 when both apply (calculated 2026-10-25).
- A dispute costs $15.00 when received and another $15.00 when countered, and the counter fee is refunded only if the dispute is won (checked 2026-10-25).
- ACH Direct Debit is 0.8% capped at $5.00 per payment, and Klarna is 5.99% + 30¢ (checked 2026-10-25).

Stripe's US rate card charges 2.9% + 30¢ per successful domestic card charge. On a $10 charge that is $0.59 and you keep $9.41; on a $100 charge it is $3.20 and you keep $96.80 (calculated 2026-10-25). Everyone quotes the percentage, but the fixed 30¢ decides whether a small ticket is worth selling at all — it is 50.8% of the fee at $10 and 9.4% at $100. This page works from the rate card as served on 2026-10-25, runs both ticket sizes, then adds the surcharges that move the number in either direction.

## What Stripe charges per card charge

| Charge | Fee at 2.9% + 30¢ | You keep | Effective rate |
|---|---|---|---|
| $10 | $0.59 | $9.41 | 5.90% |
| $20 | $0.88 | $19.12 | 4.40% |
| $50 | $1.75 | $48.25 | 3.50% |
| $100 | $3.20 | $96.80 | 3.20% |
| $250 | $7.55 | $242.45 | 3.02% |
| $1,000 | $29.30 | $970.70 | 2.93% |

[[ad]]

## What the surcharges actually mean

The headline rate is only the domestic-card case. Four published modifiers sit on top of it, and they stack.

| Scenario | Rate applied | Fee on $10 | Fee on $100 |
|---|---|---|---|
| Domestic card | 2.9% + 30¢ | $0.59 | $3.20 |
| Manually entered card | 3.4% + 30¢ | $0.64 | $3.70 |
| International card | 4.4% + 30¢ | $0.74 | $4.70 |
| Currency conversion | 3.9% + 30¢ | $0.69 | $4.20 |
| International + conversion | 5.4% + 30¢ | $0.84 | $5.70 |

The 1.5% international surcharge applies when the card is issued outside the United States; the 1% conversion surcharge applies when Stripe has to convert currency. A US merchant selling to a Canadian buyer who pays in CAD pays both, so the $100 sale nets $94.30 rather than $96.80 (calculated 2026-10-25). Manually keyed cards — phone orders, virtual terminals — add 0.5%, which exists to price the higher fraud risk of a card number typed by a human.

Disputes are priced separately from processing. Stripe lists $15.00 for a dispute received and $15.00 for a dispute countered; the counter fee comes back if you win and stays gone if you lose. On a $10 product, one lost chargeback costs $15.00 plus the refunded $10, which is 2.5× the sale price.

## How we measured it

We loaded Stripe's pricing page on 2026-10-25 with the billing country set to the United States, because the same URL otherwise serves a local-currency rate card to visitors outside it. Every rate in the tables above is printed on that page; the dollar amounts are $0.00 either way, but the units are not, and a page that quotes Hong Kong dollars is not citable as a US price. The arithmetic is ours and is labelled with the capture date. No benchmark was run and no account was opened.

## What changes in practice

The fixed fee is a floor on your pricing. At $10 a charge, 5.90% of the sale disappears; at $1,000 it is 2.93%. Any product priced under about $5 loses more than 8% to processing, which is why low-ticket sellers bundle: three $5 charges cost $1.17 in fixed fees, one $15 charge costs $0.30.

Refund behaviour is the gap worth flagging. The rate card states the fee per successful charge but does not publish whether that fee is returned when you refund the customer, so the honest planning assumption is that the 30¢ is spent the moment the charge succeeds. Budget it as a cost of doing business rather than a cost of keeping the money.

Dispute economics beat fee economics at low ticket sizes. On 1,000 sales at $10, processing takes $590; if 5% of those buyers dispute, the $15.00 fees alone add $750 — more than the entire processing bill — before any refunded principal.

## Bottom line

Price at $20 or above if the product allows it: at $20 the effective rate is 4.40%, and by $100 it is 3.20%, so the 30¢ stops being the story. If your average ticket has to stay at $10, the fixed fee is half your cost and no negotiation with Stripe changes that — only fewer, larger charges do.

**Sources:**
- Stripe pricing: https://stripe.com/pricing
- Stripe documentation: disputes: https://docs.stripe.com/disputes

---

# Stripe vs Lemon Squeezy: Fees on 1,000 Sales at $20

URL: https://actualdatabase.com/stripe-vs-lemonsqueezy-fees-1000-sales/
Verified: 2026-10-25

**Quick answer:** Stripe charges $880 in fees on 1,000 sales at $20 ($20,000 gross) and Lemon Squeezy charges $1,500. Stripe nets $19,120 at a 4.4% effective rate; Lemon Squeezy nets $18,500 at 7.5%.

- 1,000 sales at $20 gross $20,000: Stripe takes $880 in fees and Lemon Squeezy takes $1,500 (calculated 2026-10-25).
- Stripe nets $19,120 at a 4.4% effective rate; Lemon Squeezy nets $18,500 at 7.5% on the same revenue (calculated 2026-10-25).
- Lemon Squeezy costs $0.62 more per $20 sale — 2.1% of the price plus 20¢ — and that gap holds at every ticket size (calculated 2026-10-25).
- Stripe charges 2.9% + 30¢ per domestic card charge and Lemon Squeezy charges 5% + 50¢ per transaction, both with no monthly fee (checked 2026-10-25).
- At 10,000 sales the Lemon Squeezy premium reaches $6,200, which is why Lemon Squeezy offers custom pricing for high volume (calculated 2026-10-25).
- Both vendors add 1.5% for international cards, so a cross-border customer mix does not widen the gap between them (checked 2026-10-25).

The comparison that matters for a small software business is not the headline percentage but the bill at a realistic volume. At 1,000 sales of a $20 product, both vendors see $20,000 of revenue: Stripe takes $880 and Lemon Squeezy takes $1,500 (calculated 2026-10-25). That is a $620 difference for the same 1,000 customers, and it comes from two published rates rather than any negotiation. This page runs the arithmetic, then works out the volume at which Lemon Squeezy's merchant-of-record service stops being an expensive way to buy nothing.

## Fee totals at 1,000 sales of $20

| Line item | Stripe | Lemon Squeezy |
|---|---|---|
| Gross revenue | $20,000.00 | $20,000.00 |
| Percentage fee | $580.00 (2.9%) | $1,000.00 (5%) |
| Fixed fees | $300.00 (1,000 × 30¢) | $500.00 (1,000 × 50¢) |
| Total fees | $880.00 | $1,500.00 |
| Net revenue | $19,120.00 | $18,500.00 |
| Effective rate | 4.40% | 7.50% |
| Fee per sale | $0.88 | $1.50 |

[[ad]]

## What the gap actually means

The difference is not a percentage that shrinks with scale; it is a fixed formula. Stripe charges 2.9% + 30¢ and Lemon Squeezy charges 5% + 50¢, so the premium per sale is 2.1% of the price plus 20¢ — $0.62 at a $20 ticket, at 1,000 sales or at 100,000.

That has two consequences worth naming. First, the gap grows in proportion to revenue, not to customer count, so raising the price makes the premium worse and raising volume makes it worse at the same rate. Second, a foreign customer mix does not change it: both vendors add 1.5% for international cards, so with 25% of buyers outside the US the totals become $955 for Stripe and $1,575 for Lemon Squeezy, still $620 apart.

| Annual volume at $20 | Gross | Stripe fees | Lemon Squeezy fees | Premium |
|---|---|---|---|---|
| 100 sales | $2,000 | $88 | $150 | $62 |
| 1,000 sales | $20,000 | $880 | $1,500 | $620 |
| 10,000 sales | $200,000 | $8,800 | $15,000 | $6,200 |

## How we measured it

We loaded both pricing pages on 2026-10-25, plus Lemon Squeezy's fees article for the international and subscription surcharges, and multiplied the published rates by 1,000 transactions at a $20 order value. No account was opened, no test charge was made, and no benchmark was run; every figure above is either printed on a vendor page or simple multiplication of one. The international scenario assumes the same 1.5% surcharge both vendors publish.

## What changes in practice

The $620 is only a bargain if it replaces something you would otherwise buy. Lemon Squeezy is the merchant of record, which means it calculates, collects, files and remits sales tax and VAT on your behalf — for a seller with customers in the EU, the UK and a handful of US states, that is a service with a real price, and $620 a year is cheap against a tax advisor's invoice. For a seller with one country of customers and no registration obligation, it is $620 for a checkout page.

The break-even is therefore a compliance number, not a volume number. If your alternative costs $600 a year in tax tooling, the premium at $20 a ticket pays for itself at roughly 968 sales (calculated 2026-10-25); below that, Stripe plus a spreadsheet is cheaper, and above 10,000 sales Lemon Squeezy's own custom pricing exists because the formula stops being defensible.

If the $20 is a subscription rather than a one-off, Lemon Squeezy adds 0.5% on recurring charges, which widens the gap to 2.6% plus 20¢ — $0.72 per renewal (calculated 2026-10-25).

One more asymmetry is worth checking before you switch. Stripe's rate is per successful charge, so failed attempts and abandoned carts cost nothing; Lemon Squeezy's fee lands at purchase. Neither vendor's page prices a failed payment, which means the comparison above is a comparison of completed sales only.

## Bottom line

At 1,000 sales of $20, Stripe keeps $19,120 and Lemon Squeezy keeps $18,500 — a $620 premium for tax handling. Buy the premium if registering for VAT in three jurisdictions would cost more than $620 a year, which it usually does; at $20 a ticket, the crossover is about 968 sales against $600 of annual tax tooling.

**Sources:**
- Stripe pricing: https://stripe.com/pricing
- Lemon Squeezy pricing: https://www.lemonsqueezy.com/pricing
- Lemon Squeezy docs: fees: https://docs.lemonsqueezy.com/help/getting-started/fees

---

# Supabase compute instances: the full price list from Micro to 16XL

URL: https://actualdatabase.com/supabase-compute-instances-price-list/
Verified: 2026-10-25

**Quick answer:** Supabase compute runs from Micro at $10/month for 1 GB of RAM to 16XL at $3,730/month for 256 GB, and every paid plan includes $10/month of compute credit that covers one Micro instance. A medium app usually lands on Medium: $60/month for 4 GB of RAM on shared CPU, which is $50/month on top of Pro's $25 after the credit.

- Supabase compute starts at Micro: $10/month for 1 GB of RAM on shared CPU, with 60 direct connections and 200 pooler connections (checked 2026-10-25).
- The largest instance, 16XL, is $3,730/month for 64 vCPU and 256 GB of RAM, with 500 direct and 12,000 pooler connections (checked 2026-10-25).
- Large is the first dedicated instance at $110/month for 2 vCPU and 8 GB of RAM; Micro, Small and Medium run on shared CPU (checked 2026-10-25).
- Paid plans include $10/month of compute credit, enough to cover one Micro instance, and the first instance is free on paid plans (checked 2026-10-25).
- Medium costs $60/month for 4 GB of RAM, four times Small's $15/month for twice the RAM (calculated 2026-10-25).
- Compute is billed by the hour and a partial hour is still billed as a full hour; credits reset monthly and do not accumulate (checked 2026-10-25).

Supabase's plan price and its compute price are two different lines, and the second one decides the bill. The Pro plan starts at $25 a month; the instance underneath it ranges from $10 to $3,730. Every paid plan carries a $10 monthly compute credit, so the question is not whether compute is included but how far that credit stretches.

A medium-sized application typically lands on Medium: 4 GB of RAM, shared CPU, $60 a month, of which $50 is left after the credit. The steps either side of it are where most of the pricing decisions actually sit.

## The instance ladder

| Instance | RAM | CPU | Price/month | Left after $10 credit |
|---|---|---|---|---|
| Micro | 1 GB | Shared | $10 | $0 |
| Small | 2 GB | Shared | $15 | $5 |
| Medium | 4 GB | Shared | $60 | $50 |
| Large | 8 GB | 2 dedicated vCPU | $110 | $100 |
| XL | 16 GB | 4 vCPU | $210 | $200 |
| 2XL | 32 GB | 8 vCPU | $410 | $400 |
| 4XL | 64 GB | 16 vCPU | $960 | $950 |
| 8XL | 128 GB | 32 vCPU | $1,870 | $1,860 |
| 12XL | 192 GB | 48 vCPU | $2,800 | $2,790 |
| 16XL | 256 GB | 64 vCPU | $3,730 | $3,720 |

Connection limits scale with the instance: Micro allows 60 direct and 200 pooler connections, Medium 120 and 600, and 16XL 500 and 12,000. Compute is billed hourly, and the pricing page notes you can scale up or down at any time.

[[ad]]

## What the numbers actually mean

The ladder is not linear. Micro to Small is $5 for another gigabyte. Small to Medium is $45 for another two, and the price per gigabyte jumps from $7.50 on Small to $15 on Medium. That is not a rounding artefact: Medium is the first size meant for production traffic, and it is priced accordingly.

Above Medium the curve flattens again in per-gigabyte terms — Large is $13.75 per GB, XL $13.13, 2XL $12.81 — because those instances are sold on dedicated vCPU as much as on memory. 4XL reverses it at $15 per GB, which is the point where the shape of the offer changes from "database" to "dedicated server".

The credit is the one discount that applies to all of them. $10 a month covers a Micro instance exactly, and the pricing table marks the first instance as free on paid plans. It does not stack: it is one credit per organization per month, so a second project adds its own compute on top of the $10 per-project fee.

## How we measured it

Every price, RAM figure and connection count is quoted from the Supabase pricing page on 2026-10-25. The hourly billing rule, the partial-hour rounding and the statement that credits reset monthly without accumulating come from Supabase's compute usage documentation on the same date. The "left after credit" column subtracts the published $10 credit from the published instance price and is labelled calculated where it appears as a total.

## What changes in practice

The jump from Small to Medium is the single most common surprise. A side project on Pro with a Micro instance pays $25 and nothing more. The moment it needs 4 GB, the bill is $75, not $25 — three times the plan price for a change most teams make in a single click in the dashboard.

Partial hours matter less than they look but they do bite in the wrong direction. Supabase bills a full hour for any part of an hour used, so a project restarted five times in an afternoon is billed for five hours of whichever size was live. Resizing down after a load test does not refund the hours already spent at the bigger size.

The credit also hides a step. Because $10 covers Micro exactly, a team that moves to Small sees a $5 increase and concludes that compute is nearly free. Small is 2 GB of shared RAM; the next click is $45 more.

The connection counts are the second hidden step. Medium allows 120 direct and 600 pooler connections against Micro's 60 and 200, and a serverless app that opens a connection per invocation exhausts Micro's 60 long before it exhausts 1 GB of RAM. Routing through the pooler endpoint is the cheaper fix than buying a bigger instance.

## Bottom line

Budget $25 a month for Pro plus a Micro instance, and expect the real bill to be $75 once the app needs 4 GB — that is Medium at $60 minus the $10 credit, on top of the plan. Stay on Small at $15 as long as 2 GB and 90 direct connections hold; the $45 step to Medium is the one worth measuring before you take it.

**Sources:**
- Supabase pricing: https://supabase.com/pricing
- Supabase docs: manage compute usage: https://supabase.com/docs/guides/platform/manage-your-usage/compute

---

# Supabase vs Neon vs Turso: What the Free Tier Actually Includes

URL: https://actualdatabase.com/supabase-vs-neon-vs-turso-free-tier-limits/
Verified: 2026-10-25

**Quick answer:** Supabase Free gives 5 GB egress, 1 GB file storage and 2 active projects that pause after one week idle. Neon Free gives 100 CU-hours of Postgres compute and 1 GB of storage per project, across up to 100 projects. Turso Free gives 5 GB of storage with 500 million rows read and 10 million rows written per month.

- Supabase Free: 5 GB egress, 5 GB cached egress, 1 GB file storage, 2 active projects (checked 2026-10-25).
- Neon Free: 100 CU-hours per project per month, 1 GB storage per project, up to 2 CU (8 GB RAM), 5 GB object storage, up to 100 projects (checked 2026-10-25).
- Turso Free: 5 GB storage, 500 million rows read per month, 10 million rows written per month, 3 GB of syncs (checked 2026-10-25).
- Supabase Pro costs $25/month and includes 250 GB egress, 8 GB disk per project, 100 GB file storage and 100,000 MAU (checked 2026-10-25).
- Neon Launch bills $0.106 per CU-hour, $0.35 per GB-month of database storage and $0.20 per GB-month of history storage (checked 2026-10-25).
- Turso Developer costs $4.99/month with 9 GB storage, 2.5 billion rows read and 25 million rows written (checked 2026-10-25).

Three managed databases advertise a free tier that is generous enough to ship a real product on. They measure that generosity in completely different units — egress, compute hours, rows read — which makes the marketing pages hard to compare directly.

This page puts the three free tiers in one table, in the vendor's own units, with the capture date attached. Then it does the same for what happens the month you cross the line.

## What each free tier includes

| Limit | Supabase Free | Neon Free | Turso Free |
|---|---|---|---|
| Monthly price | $0 | $0 | $0 |
| Network egress | 5 GB (+5 GB cached) | Not billed on the free plan | Not billed separately |
| Storage | 1 GB file storage | 1 GB database storage per project, 5 GB object storage | 5 GB total |
| Compute allowance | Shared instance | 100 CU-hours per project, capped at 2 CU (8 GB RAM) | Rows, not compute |
| Work allowance | Not metered | 100 CU-hours per project | 500M rows read, 10M rows written, 3 GB syncs |
| Project limit | 2 active projects | Up to 100 projects | One database per plan slot |
| Idle behaviour | Free projects pause after 1 week of inactivity | Scales to zero, no CU-hours consumed | Idle costs storage only |

[[ad]]

## The unit each vendor chose tells you what it expects

Supabase meters the edge of the system — egress, file storage, monthly active users. That fits a product where the database is one part of a bundle: auth, storage, realtime, edge functions.

Neon meters compute as CU-hours, where one CU is roughly 4 GB of RAM plus the CPUs that go with it. The free plan gives each project 100 CU-hours a month and scales to zero when idle. Neon's own worked example is the honest version of this: a 0.25 CU database running 12 hours a day for 30 days uses about 90 CU-hours, so the free allowance covers a database that is awake half the day.

Turso meters the query, not the server. A free account can read 500 million rows and write 10 million rows a month whatever the CPU does, which rewards read-heavy apps with narrow rows and punishes bulk imports.

## What you pay the month you outgrow free

| Item | Supabase Pro | Neon Launch | Turso Developer |
|---|---|---|---|
| Base price | $25/month | Usage-based | $4.99/month |
| Included storage | 8 GB disk per project, 100 GB file storage | None — billed from zero | 9 GB |
| Storage overage | $0.125/GB disk, $0.0213/GB file | $0.35/GB-month | $0.75/GB |
| Compute | $10/month credits included; instances from $10 (1 GB RAM) to $3,730 (256 GB) | $0.106 per CU-hour | Rows included: 2.5B read, 25M written |
| Traffic | 250 GB egress included, then $0.09/GB | $0.10/GB after 500 GB per project | $0.35/GB of syncs after 10 GB |
| Users | 100,000 MAU included, then $0.00325/MAU | n/a | n/a |

## How we measured it

We loaded the three public pricing pages on 2026-10-25, took the numbers printed in the plan tables and the billing FAQ, and discarded anything that only appeared in a marketing headline. Where a vendor states a rule in prose rather than a table — Supabase's one-week pause, Neon's 0.25 CU default on free projects, Turso's per-row overage — we quoted the rule instead of inventing a number for it.

No benchmark was run. These are the vendors' published prices, dated, so that a later change is visible rather than silent.

## What changes in practice

The free tier is rarely what makes you upgrade. Supabase's 2-active-project cap is the constraint most teams hit first, and a paused project is a production outage that arrives without an alert. Neon's constraint is compute hours: an always-on database with scale-to-zero disabled has a floor of roughly 182.5 CU-hours a month, which is already past the 100-hour free allowance. Turso's constraint is the write budget — 10 million rows a month is generous for application traffic and small for anything that ingests events.

One more trap sits in the paid math: Supabase's $25 Pro plan includes $10 of compute credit, which covers exactly one Micro instance. A project that needs more than the shared instance starts above the headline price, not at it.

## Bottom line

For a prototype that sleeps, Neon's per-project allowance is the most forgiving. For a small app that needs auth and file storage today, Supabase's bundle saves you three integrations before you pay anything. For a read-heavy SQLite workload that never wants to think about CPU, Turso's row budget is the one that does not surprise you.

**Sources:**
- Supabase pricing: https://supabase.com/pricing
- Neon pricing: https://neon.com/pricing
- Turso pricing: https://turso.tech/pricing

---

# Turso row reads: what they cost, and what 1 billion reads a month bills

URL: https://actualdatabase.com/turso-rows-read-billing-example/
Verified: 2026-10-25

**Quick answer:** Turso bills $1 per billion row reads beyond the allowance your plan includes, and Developer at $4.99/month includes 2.5 billion of them, so an app doing 1 billion reads a month pays $4.99. On the Free plan the same app crosses the 500 million row-read allowance and its queries fail with a BLOCKED error rather than billing an overage.

- Turso Free includes 500 million row reads and 10 million row writes per month, with no overage billing (checked 2026-10-25).
- Turso Developer at $4.99/month includes 2.5 billion row reads and charges $1 per billion beyond that (checked 2026-10-25).
- Turso Scaler at $24.92/month includes 100 billion row reads and charges $0.80 per billion beyond it (checked 2026-10-25).
- Turso Pro at $416.58/month includes 250 billion row reads and charges $0.75 per billion beyond it (checked 2026-10-25).
- 1 billion row reads a month fits inside Developer's allowance, so the bill is the $4.99 plan fee plus $0 of overage (calculated 2026-10-25).
- A row read counts a row scan, not a row returned, and exceeding a plan's quota returns a BLOCKED error rather than an overage charge (checked 2026-10-25).

Turso prices reads the way a phone plan prices minutes: a large included volume, then a per-unit rate, with a hard stop on the free tier. For an app that does a billion reads a month the arithmetic is short, and the interesting part is not the price but what counts as a read.

One billion row reads a month costs $4.99, because Developer's allowance is 2.5 billion and the plan fee is the only charge. Free cannot carry the same app at all: 500 million rows read is the limit, and Turso's documentation says exceeding a quota fails the query, not the invoice.

## What a row read costs on each plan

| Plan | Rows read included | Overage rate | 1 billion reads/month |
|---|---|---|---|
| Free | 500 million | None, queries blocked | $0 but blocked at 500M |
| Developer | 2.5 billion | $1 per billion | $4.99 |
| Scaler | 100 billion | $0.80 per billion | $24.92 |
| Pro | 250 billion | $0.75 per billion | $416.58 |

The Free plan also carries 10 million row writes, 5 GB of storage and 100 databases. Developer raises storage to 9 GB at $0.75 per GB beyond, and Scaler to 24 GB at $0.50 per GB beyond.

[[ad]]

## What the row-read number actually means

A row read is a row scan, and that distinction is where budgets go wrong. Turso's documentation is explicit: the term row read refers to a row scan during statement execution, queries may scan more rows than they return, each updated row results in at least one scan, and aggregates, full table scans, joins and some ALTER TABLE statements scan a row at a time. Even `select 1` counts as one row read.

So the unit is closer to "pages examined" than "results delivered". An app that fetches 1,000 rows by primary key is cheap. The same app running a report over 1,000,000 rows without an index reads 1,000,000 rows to return a number, and repeating that report 1,000 times a month is a billion reads on its own.

The allowance structure is unusual in one respect: it is generous by an order of magnitude. Developer includes 2.5 billion reads for $4.99, which is $1.996 per billion at full consumption, and Scaler includes 100 billion for $24.92, which is $0.25 per billion. The overage rates only matter when a plan is nearly exhausted.

## How we measured it

The included volumes, overage rates and plan prices are quoted from Turso's pricing page on 2026-10-25; the definition of a row read and the BLOCKED behaviour come from Turso's usage and billing documentation on the same date. Nothing was provisioned and no query was run against a Turso database, so no scan multipliers were measured. The 1 billion-read example is arithmetic against published allowances and is marked calculated.

## What changes in practice

The first practical consequence is that a blocked query is worse than a bill. On Free, crossing 500 million reads stops the app rather than charging for the excess, so a read-heavy prototype needs Developer at $4.99 before it ships, not after it fails.

The second is that indexing is a billing decision. Every full table scan is charged per row, which means the same schema change that fixes latency also cuts the bill. A monthly report that scans 2 million rows costs 2 million reads; add the index and it reads thousands.

The third is the crossover between the two paid tiers. Developer's $1 per billion sounds cheap until volume passes 22.4 billion reads, where $4.99 plus overage exceeds Scaler's $24.92 — and Scaler also drops the rate to $0.80 per billion with far more headroom.

The fourth is write amplification on the same meter. Turso counts each updated row as at least one row read and one row write, so a nightly job that updates 1 million rows consumes 1 million of each and returns nothing — 30 million writes a month against Developer's 25 million allowance.

## Bottom line

For an app reading 1 billion rows a month, Developer at $4.99 is the whole bill and the cheapest tier that will not block. Budget Scaler's $24.92 from roughly 22 billion reads upward, and treat row scans — not row counts returned — as the number to put on a dashboard before traffic arrives.

**Sources:**
- Turso pricing: https://turso.tech/pricing
- Turso docs: usage and billing: https://docs.turso.tech/help/usage-and-billing

---

# Twilio SMS Cost per 1,000 Messages in the US

URL: https://actualdatabase.com/twilio-sms-cost-per-1000-messages/
Verified: 2026-10-25

**Quick answer:** Twilio charges $0.0083 per 160-character SMS segment in the US, and US carriers add their own per-message fee on top: $0.0035 for AT&T, $0.0045 for T-Mobile and $0.005 for Verizon. That makes 1,000 messages cost $8.30 at the base rate, $11.80 with AT&T fees or $12.80 with T-Mobile fees, and 100,000 messages cost $830, $1,180 or $1,280 on the same basis.

- Twilio's US SMS base rate is $0.0083 per 160-character segment, for outbound and inbound segments alike (checked 2026-10-25).
- US carriers add their own per-message fees: $0.0035 for AT&T, $0.0045 for T-Mobile and $0.005 for Verizon (checked 2026-10-25).
- A single outbound message therefore costs $0.0118 to AT&T subscribers and $0.0128 to T-Mobile subscribers (calculated 2026-10-25).
- 1,000 messages cost $8.30 at the base rate, $11.80 with AT&T fees and $12.80 with T-Mobile fees (calculated 2026-10-25).
- 100,000 messages cost $830, $1,180 or $1,280 on the same three bases (calculated 2026-10-25).
- MMS is a flat $0.022 per outbound message in the US, with inbound MMS at $0.0165 for long and short codes (checked 2026-10-25).

Twilio quotes one rate and bills two. The published US rate is $0.0083 per 160-character SMS segment, and US carriers add a per-message fee that Twilio passes through automatically: $0.0035 for AT&T, $0.0045 for T-Mobile and $0.005 for Verizon. Add those up and a message to an AT&T subscriber really costs $0.0118 and one to a T-Mobile subscriber $0.0128. At volume, 1,000 messages are $8.30 on the base rate but $11.80 or $12.80 once carrier fees land, and 100,000 messages are $830, $1,180 or $1,280.

## Per-message cost by destination network

| Destination network | Carrier fee | Twilio base rate | Cost per message | 1,000 messages |
| --- | --- | --- | --- | --- |
| Any US network, base only | none | $0.0083 | $0.0083 | $8.30 |
| AT&T | $0.0035 | $0.0083 | $0.0118 | $11.80 |
| T-Mobile | $0.0045 | $0.0083 | $0.0128 | $12.80 |
| Verizon | $0.005 | $0.0083 | $0.0133 | $13.30 |

Carrier fees are added to the Twilio rate (calculated 2026-10-25).

[[ad]]

## What volume does to the bill

| Volume | Base rate only | With AT&T fees | With T-Mobile fees |
| --- | --- | --- | --- |
| 1,000 messages | $8.30 | $11.80 | $12.80 |
| 10,000 messages | $83.00 | $118.00 | $128.00 |
| 100,000 messages | $830.00 | $1,180.00 | $1,280.00 |

All figures are the published per-message rates multiplied by volume (calculated 2026-10-25).

## What the numbers actually mean

There is no volume discount in the published US rate. The same $0.0083 applies at one message and at a million, so the only lever on a large bill is the mix of destinations and the number of segments per message. A 100,000-message campaign to AT&T numbers costs $1,180, and the $350 of that which is carrier fee is 30% of the total (calculated 2026-10-25).

Segments matter more than most senders expect. The rate is per 160-character segment, so a message with a link and a slightly long sentence can cross into two segments and double the cost, and three segments triple it. At 320 characters, the base rate becomes $0.0166 per message and the AT&T total $0.0236 (calculated 2026-10-25), which is the difference between a $830 campaign and a $1,660 one at 100,000 messages.

Carrier fees also mean the bill depends on who your customers are, not how many. The spread between the cheapest and most expensive listed networks is $0.0015 per message, so an audience concentrated on T-Mobile pays 8% more than one concentrated on AT&T at the same volume (calculated 2026-10-25). That is small per message and material across a year of notifications.

MMS is priced separately and flatly. Outbound MMS in the US is $0.022 per message, inbound MMS is $0.0165 for long and short codes and $0.02 for toll-free numbers, and the page does not apply the 160-character segmentation rule to MMS. Sending an image instead of text costs roughly 2.7 times the SMS base rate (calculated 2026-10-25), which is the cheapest way to deliver a long message as a single unit.

## How we measured it

Rates come from Twilio's US SMS pricing page fetched on 2026-10-25, where the base rate, the carrier fee table and the segment rule appear in the page's pricing FAQ, and the same rates are repeated on Twilio's general pricing page. No account was opened, no number was provisioned and no message was sent. Totals at volume and per-message sums are arithmetic on published rates and are marked calculated.

## What changes in practice

Budget the carrier fee, not the headline rate. A $0.0083 quote is $0.0118 to $0.0133 delivered, so any forecast built on the base rate understates the bill by 42% to 60% (calculated 2026-10-25).

Keep messages inside one segment. 160 characters including the link, at 1,000 messages, is the difference between $8.30 and $16.60 (calculated 2026-10-25), and no volume discount will recover it.

Watch the optional add-ons. The Engagement Suite and Compliance Toolkit each add $0.015 per outbound message, which is 1.8 times the base rate for either one; on 100,000 messages that is $1,500 per add-on (calculated 2026-10-25), larger than the carrier fees themselves.

## Bottom line

For 1,000 SMS messages in the US, budget $11.80 rather than $8.30, because carrier fees are charged automatically and land on every message (calculated 2026-10-25). At 100,000 messages the delivered cost is $1,180 to $1,330 depending on the network mix, and keeping to a single 160-character segment is the only saving that scales with volume.

**Sources:**
- Twilio SMS pricing, United States: https://www.twilio.com/en-us/sms/pricing/us
- Twilio pricing: https://www.twilio.com/en-us/pricing

---

# What a $10 Product Costs in Fees: Stripe vs Lemon Squeezy

URL: https://actualdatabase.com/what-a-10-dollar-product-costs-in-fees/
Verified: 2026-10-25

**Quick answer:** A $10 product nets $9.41 on Stripe at 2.9% + 30¢ and $9.00 on Lemon Squeezy at 5% + 50¢. The fixed fee is more than half of what either vendor takes at that ticket size.

- A $10 digital product nets $9.41 on Stripe (2.9% + 30¢) and $9.00 on Lemon Squeezy (5% + 50¢) (calculated 2026-10-25).
- The fixed fee is more than half of either vendor's charge on a $10 sale: 30¢ of $0.59 on Stripe and 50¢ of $1.00 on Lemon Squeezy (calculated 2026-10-25).
- To net $9.00 per unit you must price $9.58 on Stripe or $10.00 on Lemon Squeezy (calculated 2026-10-25).
- An international card takes the Stripe fee to $0.74 on a $10 charge, and a foreign-currency sale with conversion to $0.84 (calculated 2026-10-25).
- A $10 product sold into 20% VAT invoices at $12.00 on Lemon Squeezy, incurs $1.28 and leaves $8.72 after tax (calculated 2026-10-25).
- Stripe charges 2.9% + 30¢ per domestic card charge and Lemon Squeezy 5% + 50¢ per transaction, both with no monthly fee (checked 2026-10-25).

The $10 price point is where fee structures stop being comparable and start being decisive. On a domestic card, Stripe takes $0.59 and Lemon Squeezy takes $1.00 of that $10, leaving $9.41 and $9.00 respectively (calculated 2026-10-25). Neither number is large in isolation, but the fixed component behind them does not shrink with price, which is what makes a cheap product expensive to sell. This page runs the same $10 through both vendors and through the surcharges that apply outside the simplest case.

## What a $10 product nets after fees

| Scenario | Fee | You keep | Fee as % of $10 |
|---|---|---|---|
| Stripe, domestic card | $0.59 | $9.41 | 5.9% |
| Stripe, international card | $0.74 | $9.26 | 7.4% |
| Stripe, international + conversion | $0.84 | $9.16 | 8.4% |
| Lemon Squeezy, US card | $1.00 | $9.00 | 10.0% |
| Lemon Squeezy, international or PayPal | $1.15 | $8.85 | 11.5% |
| Lemon Squeezy, 20% VAT order ($12.00 total) | $1.28 | $8.72 | 10.7% of net |

[[ad]]

## What the fixed fee actually means

Stripe's 30¢ is 50.8% of the fee it charges on a $10 sale. Lemon Squeezy's 50¢ is exactly 50.0% of its $1.00. More than half of what you pay either vendor at this price point is a flat charge for the transaction existing, and no amount of volume discounting touches it.

The clean way to see the effect is to work backwards from a target. To keep $9.00 per unit you have to price $9.58 on Stripe or $10.00 on Lemon Squeezy (calculated 2026-10-25). To keep the same $9.41 that Stripe leaves you, the Lemon Squeezy price has to be $10.43. A product listed at $10 is not a $10 product on either platform once the target net is set.

## How we measured it

We loaded Stripe's pricing page and Lemon Squeezy's pricing page and fees article on 2026-10-25, took the published rates — 2.9% + 30¢, 1.5% international, 1% currency conversion, 5% + 50¢, 1.5% international or PayPal surcharge — and applied them to a $10 order, including Lemon Squeezy's own 20% VAT worked example scaled to that price. No checkout was run and no benchmark was performed; the arithmetic is ours and dated. Where a vendor does not publish a rule, such as how a refund affects the fee, we say so instead of guessing.

## What changes in practice

Small tickets are a structural disadvantage, not a negotiable one. At $10, 5.9% of a Stripe sale is 2% of pure percentage and 3.9% of flat fee; at $100 the same 30¢ is 0.3% of the sale. That is why digital products cluster at $19, $29 and $49 rather than $9 — the pricing is not psychology, it is the fixed fee being diluted. Two $5 sales cost $0.89 in fees on Stripe and $1.50 on Lemon Squeezy, against $0.59 and $1.00 for a single $10 sale, so bundling beats any plan choice at this scale (calculated 2026-10-25).

Two traps sit inside the $10 case. Currency conversion is charged on top of the international surcharge rather than instead of it, so a Canadian buyer paying in CAD costs $0.84 rather than $0.74, an eighth of the margin on a product with thin margins. And on Lemon Squeezy the fee is computed on the tax-inclusive total: adding 20% VAT raises the fee from $1.00 to $1.28 even though the extra $2.00 was never yours.

Refunds are the unresolved edge. Neither pricing page states whether the processing fee is returned when a sale is refunded, so the prudent model books the fee as spent at authorization. At $10, a 10% refund rate would otherwise look like a $0.06 problem per sale and is more likely a $0.10 one.

## Bottom line

A $10 product nets $9.41 on Stripe and $9.00 on Lemon Squeezy, and at that price the fixed fee is over half of what you pay either vendor. If the product can carry a $20 or $29 price instead, the effective Stripe rate falls to 4.4% and 3.9% respectively; if it cannot, plan on keeping 90% of revenue at best rather than the 97% the headline percentage implies.

**Sources:**
- Stripe pricing: https://stripe.com/pricing
- Lemon Squeezy pricing: https://www.lemonsqueezy.com/pricing
- Lemon Squeezy docs: fees: https://docs.lemonsqueezy.com/help/getting-started/fees

---

# How Many Page Views You Need Before Ad Money Arrives (Payout Math, Not Hype)

URL: https://actualdatabase.com/page-views-needed-to-earn/
Verified: 2026-10-25

**Quick answer:** At a $5 RPM you need about 20,000 monthly page views to clear $100, and about 200,000 to clear $1,000. RPM, not traffic, is the whole game: a $40 RPM niche needs 2,500 views for the same $100. Ad networks also hold payouts until a threshold, usually $100, so small sites can be profitable on paper and still see no money for months.

- $100 per month requires 20,000 page views at a $5 RPM, 50,000 at $2, or 2,500 at $40 (calculated 2026-10-25).
- $1,000 per month requires 200,000 page views at a $5 RPM, or 25,000 at a $40 RPM (calculated 2026-10-25).
- Google AdSense pays out at a $100 threshold, issued between the 21st and the 26th of the following month (checked 2026-10-25).
- A page view from tier-1 English traffic is typically worth 5鈥?0x a page view from tier-3 traffic in display-ad RPM terms (estimate, 2026-10-25).
- At 300 page views per article per month, clearing $100 at a $5 RPM takes about 67 published articles (calculated 2026-10-25).

## The only formula that matters

```
monthly revenue = page views / 1000 脳 RPM
page views needed = target revenue 梅 RPM 脳 1000
```

Everything else 鈥?niche, design, posting schedule 鈥?is a way of moving one of those two numbers. RPM is the one most people ignore, and it varies by a factor of fifty between niches.

## Page views required for a revenue target

| RPM | $100 / month | $500 / month | $1,000 / month | $5,000 / month |
| --- | --- | --- | --- | --- |
| $1 (tier-3 general) | 100,000 | 500,000 | 1,000,000 | 5,000,000 |
| $5 (English general) | 20,000 | 100,000 | 200,000 | 1,000,000 |
| $15 (B2B software) | 6,700 | 33,300 | 66,700 | 333,300 |
| $40 (finance, legal, insurance) | 2,500 | 12,500 | 25,000 | 125,000 |

*Rows are arithmetic, not forecasts: they assume every page view serves a paid impression. Real sites typically see 60鈥?5% of views monetised, so add 20鈥?0% to the figures.*

## Quick calculator

<form class="calc" onsubmit="return false">
  <label>Monthly page views <input id="pv" type="number" value="20000" min="0" step="1000"></label>
  <label>RPM in USD <input id="rpm" type="number" value="5" min="0.1" step="0.5"></label>
  <p class="calc-out">Estimated monthly revenue: <strong id="calc-rev">$100</strong></p>
</form>
<script>
(function () {
  var pv = document.getElementById('pv');
  var rpm = document.getElementById('rpm');
  var out = document.getElementById('calc-rev');
  if (!pv || !rpm || !out) return;
  function fmt(n) {
    return '$' + n.toLocaleString('en-US', { maximumFractionDigits: 0 });
  }
  function update() {
    var v = parseFloat(pv.value) || 0;
    var r = parseFloat(rpm.value) || 0;
    out.textContent = fmt((v / 1000) * r);
  }
  pv.addEventListener('input', update);
  rpm.addEventListener('input', update);
  update();
})();
</script>

## When the money actually arrives

Revenue is not payment. Three gates stand between a page view and your bank account:

1. **Threshold.** AdSense holds the balance until it reaches $100. At $20 per month, the first payment is five months away.
2. **Schedule.** Payments issue between the 21st and 26th of the month following the month in which the threshold was crossed, and only after tax and identity verification clear.
3. **Verification.** Address PIN by mail, tax status (a W-8BEN for non-US publishers), and a payout method. Each of these can silently hold funds for a month.

[[ad]]

## The number people forget

Page views per article. If a page earns 300 views per month, then $100 per month at a $5 RPM needs **67 pages**, not 20,000 views of willpower. Work backwards from that: pages that answer one specific question with a number in it beat pages that summarise a topic, because the specific question is what people type.

## What this model gets wrong

It assumes traffic you can actually get. A $40 RPM topic is usually a topic where the top ten results are owned by companies with a decade of domain authority, so the arithmetic says 2,500 views and reality says zero. Pick the niche where you can add a fact nobody else has published, then let the RPM follow.

**Sources:**
- Google AdSense payment thresholds and schedule: https://support.google.com/adsense/answer/1709871
- Google AdSense revenue share: https://support.google.com/adsense/answer/180195
- Ad network entry requirements: https://purpleads.io/blog/top-ad-networks-for-publishers/

---

# Free Static Hosting Limits Compared: Cloudflare Pages vs Netlify vs Vercel vs GitHub Pages

URL: https://actualdatabase.com/free-static-hosting-limits/
Verified: 2026-10-25

**Quick answer:** On our capture date, Cloudflare Pages was the only widely used free tier with unmetered bandwidth and a written commercial-use allowance. Netlify Free allowed 100 GB of bandwidth and 300 build minutes per month, Vercel Hobby allowed roughly 100 GB but prohibited commercial use, and GitHub Pages allowed 1 GB of published site with 100 GB per month of soft bandwidth and was not intended for e-commerce.

- Cloudflare Pages free tier: unmetered bandwidth and requests, 500 builds per month, 20,000 files per site, 25 MiB per file (checked 2026-10-25).
- Netlify Free: 100 GB bandwidth per month, 300 build minutes per month, 125,000 serverless function invocations (checked 2026-10-25).
- Vercel Hobby: about 100 GB bandwidth per month, commercial use prohibited by the plan terms (checked 2026-10-25).
- GitHub Pages: 1 GB published site size, 100 GB per month soft bandwidth, 10 builds per hour, not intended for commercial e-commerce (checked 2026-10-25).
- Only Cloudflare Pages among these four states a commercial-use allowance that does not depend on a paid plan (checked 2026-10-25).

## Why this table exists

Hosting quota pages are long, undated, and written for lawyers. The question people actually ask is narrower: *can I put ads on it, and will it fall over at 50,000 page views?* This page answers only that, and prints the date each figure was confirmed so you can tell whether it still holds.

## The comparison

| Host | Bandwidth | Builds | Commercial use | Custom domain | Notable cap |
| --- | --- | --- | --- | --- | --- |
| Cloudflare Pages | Unmetered | 500 / month | Allowed | Yes, free | 20,000 files, 25 MiB per file |
| Netlify Free | 100 GB / month | 300 minutes / month | Allowed | Yes, free | 125k function invocations |
| Vercel Hobby | ~100 GB / month | Included | Prohibited on Hobby | Yes, free | Fair-use enforcement |
| GitHub Pages | 100 GB / month soft | 10 builds / hour | Not intended for e-commerce | Yes, free | 1 GB published site |

*Verified 2026-10-25 against each vendor's own limits page. See [Sources](#sources).*

## What actually breaks first

Bandwidth is almost never the binding constraint for a text site. A 60 KB HTML page with a 20 KB stylesheet serves **12,000 page views per GB**, so Netlify's 100 GB is roughly 1.2 million views. What breaks first, in order:

1. **Build minutes** on sites that rebuild thousands of pages on every commit.
2. **File count caps** when a generator emits one file per tag, per author, per pagination page.
3. **Plan terms**, not quotas. A site that gets popular on a plan that forbids commercial use gets suspended for the wrong reason.

## The trap that costs people their ad account

AdSense verifies site ownership by asking you to place a record on the domain. That works on `example.com` and fails on `yourproject.pages.dev`, because you do not control the apex record of a domain you do not own. Buy the domain first (about $10 per year), attach it to the host, then apply.

## Decision rule

- Text, tools, docs, data pages 鈫?**Cloudflare Pages**.
- Needs serverless functions or form handling 鈫?**Netlify**, until build minutes become the constraint.
- Next.js with a commercial plan 鈫?**Vercel paid**, because Hobby forbids the commercial part.
- Repository documentation only 鈫?**GitHub Pages**.

[[ad]]

## How to keep this table honest

Run the same four pages through a diff weekly: fetch each limits page, hash the body, and alert on change. Re-verify any row that changed, update the date, and note what changed. A table that is silently six months stale is worse than no table, because people cite it.

**Sources:**
- Cloudflare Pages limits: https://developers.cloudflare.com/pages/platform/limits/
- Netlify pricing and free tier: https://www.netlify.com/pricing/
- Vercel pricing and fair use: https://vercel.com/pricing
- GitHub Pages usage limits: https://docs.github.com/en/pages/getting-started-with-github-pages/github-pages-limits