Blog · Flat-price hosting

PaaS with no usage-based billing in 2026: flat prices, spending caps and overage compared

By the Pethost team · 17 September 2026 · Updated 5 October 2026 · 10 min read

You are afraid of a surprise bill and want a PaaS with a predictable monthly price: flat-rate pricing, no usage-based billing, no overage charges. In October 2026 most platforms still meter something: CPU and memory by the second, or traffic by the gigabyte. Of six we read that bill by use or per service, only Railway documents a hard limit that stops the bill. On Pethost nothing is metered: €9.99 a month rents a whole machine, and when it is full an app slows down and the bill stays €9.99.

Where a surprise bill comes from

A surprise bill is a meter multiplied by something nobody planned.

A traffic spike

A post reaches a front page: a million views of a 500 KB page are 500 GB of outgoing traffic. For one small service, at each vendor's listed rate and allowance, by our sums:

The amounts are small, and on most of these platforms nothing stops them at ten times the traffic.

A loop

A worker retries a failing job without a pause and keeps a CPU core busy. Where compute is billed by use, that is on the invoice: on Railway one busy vCPU is $20 a month, and a leak that grows a service from 0.5 GB to 4 GB turns $5 of memory into $40. Where an instance has a fixed price, a loop adds nothing. On Pethost it takes CPU from your other projects, and the bill does not move.

A crawler

A crawler is traffic you did not ask for, billed like a spike, and on Google Cloud Run by request too: $0.40 a million. On Pethost it is a line in the request log.

An agent that deploys ten things

Where every service has its own price, each deploy is a new line: $7 for 512 MB on Render and on Heroku, $5 on DigitalOcean App Platform. Ten are $70, $70 and $50 a month. On Pethost a new project takes memory from the same machine and adds nothing to the bill, and your AI coding agent cannot pay for anything or change your account.

Spending caps in 2026: Railway's hard usage limit, Fly.io with no spending cap

When a month goes wrong on a metered platform, the service stops or the bill grows. A hard cap chooses the first.

Railway: a hard usage limit from $10

Railway has a hard limit on the whole bill, on every plan. You set an amount of $10 or more, and when usage reaches it Railway takes all your workloads offline: its documentation calls setting one a possibly destructive action.

Fly.io: no spending cap, no billing alerts

Fly.io's cost page is plain: “there's no soft ceiling. If you go over, we'll bill you.” It has no billing alerts yet. A Machine does have a fixed price for its size while it runs, $3.69 a month for 512 MB in Ashburn, whatever its CPU does. What moves is traffic, volumes, and Machines somebody forgot.

Every platform's meter, overage rate and cap

Billing by use is built for two cases: one thing that mostly idles, which then costs close to nothing, and one that must scale through a spike. Google Cloud Run scales a quiet service to zero, at the price of a cold start, and Railway's Pro plan runs a service on up to 42 replicas. Five always-on projects are neither: they hold their memory all month, and the meter counts all of it.

Render, DigitalOcean App Platform and Heroku: a fixed instance price, and bandwidth overage per GB

On these three an instance costs the same whatever it uses. What moves the bill is traffic, and the count of services.

• Render includes 5 GB of outgoing traffic a month on Hobby and 25 GB on Pro, and bills $0.15 for each gigabyte after it.
• DigitalOcean App Platform pools 50 GiB a container across the team and bills $0.02 a GiB past the pool. Its documentation says usage cannot be viewed.
• Heroku bills no traffic: a Basic dyno is a flat $7 a month, and a second app is a second dyno. See Heroku alternatives.

Flat-rate pricing with no overage charges: the machine is the limit

A flat price moves the limit from the invoice to the machine. On Pethost €9.99 a month rents a virtual machine of 2 vCPU, 3 GB of memory and 30 GB of SSD, with as many projects, containers and databases as fit. Nothing is metered, nothing is billed on top. At the limit, this happens in place of a charge:

• The CPU is busy: requests wait longer, for as long as the spike lasts.
• A service takes more memory than there is: it is killed, and the panel and your agent report it as out of memory.
• A deploy does not fit: it fails and says why: the disk is full, or how much memory the build lacks.

The bill is €9.99 in each case. Grow, a machine twice the size, is €17.99; moving up from one plan to another follows soon.

Two other hosts we read meter nothing. Sliplane is flat per server, as we are: €9 a month is 1 vCPU and 1 GB there, and the 4 GB server that holds the workload below is €28.80 (Sliplane alternatives). Hostim.dev is flat per app: €2.50 a month for 1 GB, and each further app is one more plan. Unlimited projects for one flat monthly fee compares what one price holds.

What if I just rent a VPS? Its price is fixed too

It is, and two things are worth knowing. An agent on a bare server works as root over SSH, where one wrong command takes down the proxy or a database; on Pethost its shell is inside your containers, never on the machine. And a server is not sure to cost less: Hetzner, the one most guides name, raised its prices twice in 2026, and its cheap Cost-Optimized range has long been out of stock. Those servers are hardware it bought at good prices, before the RAM shortage, and on 5 October every plan of the range was marked “not available”. The 2 vCPU, 4 GB server it does sell, CPX22, is €19.49 a month, up from €7.99, before a panel, backup storage and your hours.

What a predictable monthly price comes to for three apps, a worker and a Postgres

One workload on every host here: three always-on web services of 512 MB each, a worker, a Postgres and 10 GB of outgoing traffic a month. That is about 2.5 GB of Starter's 3 GB.

Railway's figure counts memory only, on our assumption of 0.5 GB a service: each 0.1 vCPU a service uses adds $2. Fly.io's are Ashburn prices, the lower one with a Postgres you run yourself: its managed one starts at $38. Cloud Run's is at us-central1 rates, with Cloud SQL. Railway vs Render vs Fly.io pricing has the plans line by line.

Postgres on Pethost is a container in your project, in the nightly backup: not a managed database with replicas or point-in-time recovery, as several of those priced above are.

What €9.99 a month buys when nothing is metered

Starter is 3 GB of memory that is always on. The same 3 GB at each platform's own plan prices, by our sums:

That is the best value for a fixed bill, and the work around the memory is done too. In the €9.99, without VAT:

• €0 billed on top, in any month
• 30 days of request logs
• 14 nights of encrypted backups, off the machine
• 13 MCP tools for your agent

Also inside: HTTPS for every project, deploys from GitHub, SSH into any container. On Render, HTTP request logs need the Pro workspace, $25 a month more, and DigitalOcean's $7 database is not backed up by default. There is no free tier: €9.99 rents a real machine, so nothing sleeps. See the plans.

Who Pethost is for

Read from each vendor's own pricing page and documentation on 5 October 2026. Prices are in each vendor's currency; ours are without VAT.

1. Hetzner Cloud: Cost-Optimized plans
2. Hetzner: price adjustment
3. Railway: pricing
4. Railway: cost control
5. Render: pricing
6. Render: outbound bandwidth
7. Render: the plans of April 2026
8. Fly.io: pricing
9. Fly.io: cost management
10. Fly.io: billing
11. Heroku: pricing
12. Heroku: usage and billing
13. Heroku: limits
14. Heroku: an update
15. DigitalOcean App Platform: pricing
16. DigitalOcean App Platform: billing
17. DigitalOcean App Platform: limits
18. Google Cloud Run: pricing
19. Google Cloud: network pricing
20. Google Cloud: budgets
21. Vercel: pricing
22. Netlify: pricing
23. Sliplane: pricing
24. Sliplane: fair use
25. Hostim.dev: pricing