Updated Aug 20, 2026

Self-Hosted AI

Running models on your own hardware instead of calling a vendor API — for control and data residency, at the cost of running it yourself.

Share

What it means

Self-hosting means taking an open-weight model and running it on infrastructure you control: your servers, your cloud tenancy, or a workstation. Nothing leaves your environment.

The reasons are usually data residency and control rather than cost. Regulated data that cannot go to a third party, air-gapped environments, and a desire not to have capability or pricing change beneath you are the durable motivations. Cost only favors self-hosting at sustained high volume, because you pay for hardware whether it is busy or idle, while an API charges only for use.

What gets underestimated is operational burden: capacity planning, GPU availability, model updates, monitoring, and the specialist expertise to make inference efficient. Quantization is what makes it tractable at all on modest hardware.

A middle path — a vendor's model deployed inside your own cloud tenancy — covers many data-residency requirements without full self-management.

Why it matters

Self-hosting is the answer to a specific set of constraints, and a common misdiagnosis for cost. Teams that switch to save money frequently discover the engineering time exceeds the API bill they were avoiding.

In practice

Decide on data-residency and control grounds first, cost second. Read the actual license before assuming an "open" model permits your commercial use, and check the tenancy-hosted option before committing to run infrastructure.

Where this shows up

Tools and models in our catalog.

Related terms