On-Premise AI
Also known as: Local AI · Self-hosted AI
On-premise AI means running a model on hardware you control instead of sending requests to an external AI provider.
In most AI applications, requests travel over the internet to a model running on an external provider's servers. On-premise AI instead means running a model on hardware owned or controlled by the business itself, so requests and data never have to leave the company network. The usual reason is data sovereignty: where the data physically sits, and who else has access to it.
In practice this requires a suitable, usually openly available model, servers powerful enough with an appropriate graphics card, and ongoing operation of that infrastructure, similar to running your own file server. The models used are typically smaller and noticeably behind the largest models from major providers in capability, since their scale of compute is hard to match in-house.
One limiting case is assuming on-premise AI is automatically safer. The data may not leave the company, but the server itself still has to be secured, patched and monitored like any other system — otherwise the risk simply moves elsewhere.
What it means in practice
For a business of around fifty people, on-premise AI mainly pays off when genuinely sensitive customer or production data needs processing that, for real reasons, may not leave the company network, and when there's enough ongoing demand to justify the hardware and its upkeep. For occasional use, an external provider is usually the more practical route. NDVDL advises on this trade-off case by case.
Is this handled properly at your site?
We look at how it actually stands with you — and say honestly whether anything needs doing.
Related terms
All termsA term from your quote missing here?
Send us the passage you do not follow. We will explain it — with no obligation to order anything.