Priced against a PaaS
On a PaaS the web process, the queue worker, the database and staging are each a line on the bill. On your own server they share one machine, and Provisore charges for managing it. One app with a worker and Postgres comes to $22.10 a month here, machine included, and $38 on the cheapest platform below. Every line links to its rate card. Prices checked 12 September 2026.
Two bills, or one
What the meter is attached to decides how the bill grows.
- A PaaS bills per piece
- The web process is one line. The queue worker is another. The database is a managed add-on with its own rate card, and staging is a second copy of all three. The platform runs them for you, which is worth paying for, and the bill grows with the number of pieces rather than with the traffic they serve.
- A server bills per machine
- One machine runs the web process, the worker, Postgres, Redis, the cron entries and staging. It costs the same whether you run one of those or all six, and the bill grows only when you outgrow the hardware.
- Provisore bills for the part it runs
- Your provider invoices you for the machine and Provisore invoices you for managing it. Deploys, databases, daemons, SSL and the firewall run on hardware you already pay for, so every plan includes them. The plans meter the monitoring, which runs on our servers.
One app in production
A web application with a background queue worker and a Postgres database. No staging, nothing clever.
- A web process
- One queue worker
- A Postgres database
Heroku
- Standard-1X web dyno$25
- Standard-1X worker dyno$25
- Postgres Essential-2$20
Render
- Standard web service$25
- Starter background worker$7
- Postgres Basic$6
Fly.io
- shared-cpu-4x, 1 GB$7.78
- shared-cpu-2x, 512 MB$3.89
- Managed Postgres Basic$38
Running Postgres yourself on a second Fly machine costs a fraction of the managed cluster, and gives up its failover and backups.
The same app, with staging
The app above, plus a staging copy with its own database. Here the two billing shapes separate.
- A web process
- One queue worker
- A Postgres database
- A staging web process
- A staging database
Provisore Solo
- Provisore Solo$12
- netcup VPS 1000 G12$10.10
Staging is another site and another database on the machine already paid for, so neither line moves.
Heroku
- Standard-1X web dyno$25
- Standard-1X worker dyno$25
- Postgres Essential-2$20
- Basic dyno, staging$7
- Postgres Essential-0, staging$5
Render
- Standard web service$25
- Starter background worker$7
- Postgres Basic$6
- Starter web service, staging$7
- Postgres Basic, staging$6
Fly.io
- shared-cpu-4x, 1 GB$7.78
- shared-cpu-2x, 512 MB$3.89
- Managed Postgres Basic$38
- shared-cpu-1x, 256 MB, staging$1.94
- Managed Postgres Basic, staging$38
Two managed Postgres clusters are most of this total. A staging database on a plain machine would cut it by about a third.
Twelve client sites
An agency running twelve sites, each with its own database, four of them with a queue worker. Two machines, because one estate should not sit on one box.
- 12 web processes
- 12 databases
- 4 queue workers
- Room to lose a machine
Provisore Pro
- Provisore Pro$29
- 2 x netcup VPS 2000 G12$37.54
Pro carries unlimited connected servers and unlimited sites, so the estate grows by adding sites rather than by adding invoices.
Heroku
- 12 x Basic dyno$84
- 12 x Postgres Essential-1$108
- 4 x Basic worker dyno$28
Render
- 12 x Starter web service$84
- 12 x Postgres Basic$72
- 4 x Starter background worker$28
Fly.io
- 12 x shared-cpu-2x, 512 MB$46.68
- 12 x Managed Postgres Basic$456
- 4 x shared-cpu-1x, 256 MB$7.76
Twelve managed clusters is the expensive way to read this. Most Fly estates this size run Postgres on their own machines instead.
The platforms that bill by usage
Railway and Laravel Cloud bill for the resources an application uses, so the total depends on your traffic. These are their rates.
Railway
- Base
- $5/month Hobby, $20/month Pro, each including the same value in usage
- Usage
- about $20 per vCPU per month and $10 per GB of memory per month, billed by the second
An app that idles most of the day costs a fraction of an always-on server. One that runs flat out at 1 vCPU and 2 GB costs about $40 a month before the plan fee.
Laravel Cloud
- Base
- $5/month Starter, $20/month Growth, $200/month Business, each plus usage
- Usage
- compute charged only while the app is awake, with serverless Postgres and managed queues billed separately
Scale-to-zero is a real advantage. An app used one evening a week costs close to nothing there and a full month of a server here. In exchange, the compute, the database and the queues are theirs, and they go when the subscription does.
What the totals leave out
Four things a platform gives you that a single machine does not, and two numbers moving under all of it.
Autoscaling and scale-to-zero
A machine runs at one size until you change it. A platform adds replicas under load, removes them after, and some stop charging entirely while an application is idle. If your traffic arrives in spikes, that is worth money these tables do not credit them with.
An edge network in front
CDN, edge caching and DDoS mitigation come with most platform plans. One server in one data centre has none of that, and putting a CDN in front of it is a decision and an invoice of your own.
Managed database operations
A managed Postgres does failover and point-in-time recovery. Provisore takes scheduled backups to an S3 bucket you control, from Pro up. You can restore without us, and running the restore is yours.
Somebody else's capacity planning
Deciding the box is full is your job. Every plan shows CPU, memory and disk for each server on Today, and alert rules that notify you start on Pro. Resizing the machine is yours to do.
VAT
netcup lists prices with VAT. The netcup VPS 1000 G12 lists at EUR 10.37, which is EUR 8.71 plus 19%. The platform prices here are net and taxed at checkout, so the machine lines use net too. If you cannot reclaim VAT, add it back to both sides.
The exchange rate
netcup invoices in euro and our plans are in dollars. The machine lines were converted at 1.16 on 12 September 2026. Your netcup invoice is in euro at that day's rate.
Machine prices from netcup’s own VPS page, net of VAT.
The other tools that manage your own server
Forge and Ploi also manage a server you own, and they are priced like Provisore rather than like a platform. On price alone there is little in it.
| Tool | Entry | Middle | Top |
|---|---|---|---|
| Provisore | Solo $12/month | Pro $29/month | Scale $59/month |
| Laravel Forge | Hobby $12/month | Growth $19/month | Business $39/month |
| Ploi | Basic $10/month | Pro $16/month | Unlimited $36/month |
What $12 buys here is the full server management and server metrics. The tiers above add alert rules, uptime monitors, log shipping with fleet search and on-call rotations. Provisore also builds and runs Swift services, such as Vapor and Hummingbird apps, next to PHP and Node. The full plan comparison lists every line.
Questions people ask about these numbers
Are these prices current?
Every third-party figure on this page was read from the page it links to on 12 September 2026. That date is when somebody last looked, not a promise that nothing has moved since. Follow the link in any column before you make a decision on it, and tell us if you find one that has drifted.
Why netcup?
It is cheap and its prices are public. Provisore manages Ubuntu servers you own on any provider, so buy the machine wherever you like and swap in its price. Our half of the total does not move. Provisore can also create the server for you on Hetzner, the one provider it provisions directly.
Is a PaaS ever cheaper?
Yes. An application idle most of the week costs almost nothing on a platform that scales to zero, and a full month of a server here, because a server you own runs whether anybody visits or not. If your traffic is occasional, buy the platform.
Do the totals include your time?
No, and neither do theirs. Provisore does the provisioning, the unattended security updates and the certificates, and from Pro up the alerting. It does not do capacity planning. Deciding the box has run out of memory is your call.
What about bandwidth?
The netcup plans quoted here include traffic, with a fair-use throttle rather than an overage invoice. Most platform bills meter egress separately once you pass an included allowance, so a download-heavy or media-heavy application widens the gap in a direction these tables do not show.
Can I run the comparison on my own numbers?
Start the trial and connect one server. Today shows what it runs, and Settings → Billing shows what Provisore charges for it.
Price it on your own servers.
Fourteen days on any plan, card required, and nothing on your machines changes if you walk away.