# The VPS costs less. Your time is the second bill

A self-managed VPS is the cheapest way to get a whole virtual machine. Postgres, Redis and a container runtime are free. If you already enjoy operating servers, a VPS can be the correct answer.

Lathe is not trying to beat the raw server price. It charges for the machine plus the repeatable work around it: provisioning, TLS, credentials, monitoring, patching, backups, restore checks and a control surface for an agent or a person.

![A self-managed VPS requires setup, monthly maintenance and failure response; Lathe bundles setup and ongoing operations into one managed machine.](https://docs.lathe.live/static/guides/lathe-vs-vps-cost.png)

## A quotable comparison

| Cost or task | Self-managed VPS | Lathe |
| --- | --- | --- |
| Monthly infrastructure | Your provider's server, disk, snapshots and traffic charges | $15, $29, $49 or $89 flat |
| Initial setup | Install, harden, firewall, TLS, database, pooling, backups, monitoring | About five minutes to a running machine on average |
| App deployment | Build and maintain your own path | Apps engine and deployment jobs included |
| Postgres | Install, configure, patch and monitor | Postgres 17, TLS, SCRAM and PgBouncer included |
| Redis / queue | Install, secure, persist and back up | Redis 8, CouchDB 3 and NATS 2 optional at no added price |
| Backups | Design, schedule and store them | Daily, seven kept |
| Restore proof | Build and run the test yourself | Latest backup restored to a temporary machine and checked monthly |
| Certificates | Configure and renew | Managed TLS |
| Security updates | Track and schedule | Fixed weekly maintenance window |
| Failure response | You | Automatic recovery first, founder support when it does not recover |
| Root access | Yes | No |

The table does not assign a fake hourly rate to the founder. Put in your own.

If setup takes six hours and steady operations take two hours a month, multiply those hours by what an uninterrupted product hour is worth to you. Add that to the VPS, snapshot, storage and traffic bill. If your time is effectively free and the work is enjoyable, DIY wins. If those hours displace customer work, the raw server price is not the total.

## The work people forget to price

The first install is rarely the expensive part. The recurring list is:

- patch the operating system and every data service;
- renew and debug certificates;
- watch disk, memory and connections;
- rotate credentials without breaking the app;
- verify that backups restore rather than only checking that a job exited zero;
- respond when the machine stops answering;
- document enough of it that the next change is safe.

A backup that has never been restored is a hope, not recovery evidence.

## What Lathe does not buy

Lathe is one managed machine, not an HA platform. It has no root or SSH for customers, no automatic failover, no horizontal scaling, no read replicas and no point-in-time recovery. A workload that needs those things should use a different architecture, whether managed or self-built.

It also does not make a very small VPS economically irrational. If you can run the stack safely for a few dollars and want the control, keep doing it.

## A fair decision rule

Choose a VPS when root access, custom system packages and maximum control matter, and someone genuinely owns operations.

Choose Lathe when the app fits on one machine, standard containers and protocols are enough, and the person building the product should not also be the on-call infrastructure team.

The exit is ordinary: deploy the same container elsewhere, use `pg_dump` and `pg_restore`, and move the connection string. Test that exit before production. A managed service is easier to trust when leaving it is boring.