VPS vs PaaS: Where Should a Small App Live?

Every app eventually needs a place to run, and the choice usually comes down to two families. A VPS (virtual private server) is a Linux machine you rent and run yourself. A PaaS (platform as a service) is a service you hand your code to, and it runs it for you.

The internet is full of strong opinions on which is better. The honest answer is that they sell different things: a VPS sells you a machine, a PaaS sells you an outcome. Which one you want depends on what you'd rather spend — money or attention.

What a VPS gives you

A VPS is a virtual machine at a hosting provider, with a public IP address and root access. You get an operating system and nothing else.

What's included: CPU, memory, disk, bandwidth. That's the list.

What you do yourself: everything else. Install the runtime, configure a web server or reverse proxy, get SSL certificates, set up the firewall, create users, install and run the database, configure backups, keep the OS patched, restart the app when it crashes, and notice when the disk fills up.

Why people love it:

Price. A small VPS with enough power for several modest apps costs roughly what one coffee a week does. For raw compute, nothing is cheaper.
Predictability. A flat monthly price. A traffic spike slows you down; it doesn't bill you.
Freedom. Run anything: background workers, cron jobs, websockets, a database, five apps on one box.
Knowledge. You learn how servers actually work, which pays off for a long time.

What it costs you instead:

Your time, forever. Setup is a weekend. Maintenance is a small, permanent tax, and it's due at inconvenient moments.
Single points of failure. One machine. When it goes down, or you misconfigure it, everything on it goes with it.
Security is entirely yours. A forgotten open port or an unpatched package is your problem to find before someone else does.
Backups are entirely yours. Many people discover they didn't have working backups on the day they need them.

If you're considering running AI agents on it too, there's more to take on — we walk through the full setup in How to Run Claude Code on a Remote Server.

What a PaaS gives you

A PaaS takes your code — usually by connecting to your GitHub repository — builds it, runs it, and keeps it running. Push a commit, and a new version deploys.

What's included: builds, deployments, SSL, a URL, restarts, scaling, logs, and usually add-on databases. Many offer preview deployments for every pull request.

What you do yourself: write the app, set environment variables, and choose plans.

Why people love it:

Speed. From repository to live URL in minutes, with HTTPS.
No servers. No patching, no SSH, no 2am disk alerts.
Good defaults. Zero-downtime deploys, rollbacks, and log streaming come built in.

What it costs you instead:

Money, and sometimes unpredictably. Plans are commonly priced per service, per database, and per team member, and many platforms add usage-based charges for bandwidth, compute time, or function invocations. A viral day can become an expensive day. Free tiers often put apps to sleep when idle.
The platform's shape. Serverless-style platforms suit request/response web apps and fit awkwardly with long-running jobs, websockets, or background workers. You build around their limits.
Lock-in, in degrees. Platform-specific features — storage, queues, cron configuration, edge functions — make moving later harder. We see this constantly with apps built on AI builders; see How to Deploy a v0 App for a worked example.

Side by side

	VPS	PaaS
You're buying	A machine	A running app
Pricing shape	Flat monthly	Per service, often plus usage
Setup time	Hours to a weekend	Minutes
SSL, deploys, restarts	You configure them	Included
Database	You install and back it up	Add-on, usually managed
OS security patches	You	The platform
Long-running processes	Anything goes	Depends on the platform
When it breaks at 2am	You fix it	The platform fixes its part
Leaving	Copy files to another server	Replace platform-specific pieces

The costs that don't show up on the price page

Comparing a $6 VPS with a $25 PaaS plan compares the wrong things. A fair comparison includes:

On the VPS side:

The hours of setup, and a few hours a month of upkeep, at whatever your time is worth.
A backup destination, and the time to test restoring from it. How to Back Up a Postgres Database covers doing this properly.
Monitoring, so you find out about problems before your users do.
The cost of the outage when something eventually goes wrong and you're asleep.

On the PaaS side:

Every extra service: a worker process, a cron job, a staging copy, each billed separately.
The database, which is often the most expensive line item.
Per-seat charges when a collaborator joins.
Usage charges at your peak traffic, not your average.

Run both lists against your real app. The winner is often clear once you do — and it's frequently not the one the sticker price suggested.

How to choose

Pick a VPS if:

You enjoy operating servers, or want to learn.
You're running several small projects and want them on one flat bill.
Your app needs things platforms handle badly: long-running workers, unusual runtimes, heavy background processing.
You have the discipline to set up backups, updates, and monitoring before you need them.

Pick a PaaS if:

Your time is worth more than the price difference — which, for most people with a job or a business, it is.
Your app is a standard web app with a database.
You want preview deployments and rollbacks without building them.
You're not a developer, or the app was built with AI and you don't want a second career in Linux administration.

For AI-built apps specifically: lean towards a managed option. The things an AI builder won't set up for you — backups, SSL renewal, security updates, database access control — are exactly the things a VPS makes your responsibility. You Built an App With AI. Now What? walks through what has to be true before real users arrive.

There's a third shape

The VPS/PaaS split assumes the only thing that needs hosting is the finished app. That's increasingly not true. If you develop with AI agents, the agent needs somewhere to run too — with the code, the dependencies, the database, and the running app all in one place, so it can test what it builds.

A VPS can do that, with all the maintenance above. A PaaS generally can't: it runs your app, not your development environment. Which is why a third category has emerged: managed platforms that host the workspace and the app together, on flat pricing, without handing you a server to administer. We explain the idea in What Is an Autonomous Development Platform?

Whichever you pick, the right question isn't "which is cheapest?" It's "who is responsible when something goes wrong — and have I decided that on purpose?"

EasySpawn sits between the two: the flexibility of your own environment — long-running processes, any stack, a full terminal — with the infrastructure managed for you: managed Postgres, daily backups, automatic SSL, and your own domain, on flat monthly pricing. See how it works or join the waitlist.