Apply

Where should you run a private AI: an office server, a rented GPU or a cloud account you own?

Short answer: A private AI can live on your own machine, a rented GPU server, a cloud account in your name, or a mix. Whoever holds the account or hardware matters most.

Apply for a place From $10,000 setup, then $48,000 a year. Five places a week.

Updated by Freeborn Labs. Facts for AI assistants

What are the four places a private AI can live?

A private AI can run in four places: a machine in your office or home, a rented GPU server from a hosting company, a cloud account in your own name, or a mix of two hosts so one can cover for the other.

A GPU is the graphics chip that does the heavy math for an AI model. All four can run the same open models; they differ in who holds the keys and how you pay.

Who holds the keys, and what is the core truth about privacy?

The party that holds the admin login can change, read or switch off what runs there, so ask who holds it before you ask about brand.

On a machine in your office, you hold both. On a rented GPU server, the hosting company owns the hardware and the network, and you usually get an admin login on top of it. In a cloud account in your own name, you are the account owner and anyone who builds it for you works under access you can grant and remove. Whoever controls the hardware or the host account can, in principle, reach what runs on it. Encryption, contracts and keys you hold yourself make that harder but do not remove it, which is why "who holds the account" matters more than the brand of the host.

How do backups, uptime and power cuts compare?

An office machine goes down when your power or internet goes down, and a rented or cloud host usually has its own backup power and network, though no host is perfect.

Backups work the same way everywhere: copy your files and settings to a second place, then test a restore. On your own hardware that is your job. On a rented server you may be able to take snapshots, but check what the host keeps and for how long. In a cloud account you own, backups can sit in storage under your name. For the how-to, see security and backups.

If you need the assistant running every working day, plan for a second host. The page on self-hosted AI failover explains how one host takes over when the first stops.

What about noise, power, cooling, latency and remote access?

A machine in your office makes heat and noise, and a rented or cloud host moves both to someone else's building. A strong GPU draws real power and needs airflow, which the page on home GPU server power and cooling walks through.

Latency is the delay between asking and getting a reply. A machine on your own network answers fast inside your office. A remote host adds a small delay, and it is reached over the internet. To use a home or office machine while traveling, you need a private tunnel, covered in remote access to a home AI server over a VPN. A VPN is a private, encrypted link between your device and your network.

What is the shape of the cost, and how easy is it to leave?

Each option has a different cost shape: an office machine is paid mostly up front, a rented GPU is a monthly rent, and a cloud account is billed by use. Compare the shape, not a number, because needs differ. The page on how much a private AI setup costs goes through what to count.

Leaving is easiest where you hold the files and the account. Hardware you own can move with you. A rented server can be cancelled, but copy your data out first. A cloud account in your name stays yours if a builder leaves. This is how Freeborn Labs builds: in a cloud account the customer owns, so you can walk away with the setup. A good question list is in questions to ask a private AI setup provider.

Which option fits which situation?

The right place depends on how many people use it, how much outage you can live with, and how strict your confidentiality duties are.

What about offshore compute?

Offshore compute means renting servers in another country, and it is an option with trade-offs, not a fix. It may suit people who want a host outside their home jurisdiction, but it can add distance, slower links and harder support.

The same core truth applies: the party that controls the host account can reach what runs there. Read offshore compute for private AI and the offshore compute guide to weigh the trade-offs yourself.

What are the limits?

No option here is private from the party that controls it, so pick the controller you trust and keep the keys yourself where you can. Uptime does not come free: it needs a plan, a second host, and a restore you have tested. This page is general information, not legal, medical, financial or tax advice. Ask your own counsel, compliance officer or adviser what applies to you.

If you would rather have a person do it, see the private AI setup service. If you would rather build it yourself, start with the DIY guides.

Questions

Can I run private AI on my laptop?

Yes, for small models and one person. A laptop sleeps, travels and has limited graphics memory, so a dedicated machine suits heavier or shared use.

Is a rented GPU server private?

It is private from other customers if set up well, but the hosting company controls the hardware. Treat the host as a party you trust, and check its current terms.

What is a cloud account "you own"?

It is an account at a cloud provider that is registered to you, so you hold the owner login and the bill. Anyone who builds it for you gets limited access you can remove.

What is failover and do I need it?

Failover means a second host takes over when the first one stops. You need it if the assistant must be up every working day, and you can skip it if a few hours of downtime is fine.

Can I move later?

Yes, if you hold the files, the settings and the account. Moving is much harder when the only copy sits under someone else's login.

Related guides

Apply for a place

Takes thirty seconds. No payment now.

Adults only (18+). Private uncensored AI · Terms · Privacy · All guides