Guide
How much does an on-prem AI system cost?
Fixed hardware cost versus per-token API billing, and where owning your AI becomes cheaper than renting.
By James Drayson
In short
An on-prem AI system is a fixed capital cost, hardware plus model plus setup, rather than a per-token bill that grows with every query. For sustained enterprise usage the payback against API fees typically lands within months, after which owning is dramatically cheaper and you keep the asset.
What you're paying for
- Hardware: GPU server(s) sized to your model, a fixed, one-time cost.
- The model: Open-weight models you own outright, with no licence to renew.
- Serving & apps: API, chat, and usage platform.
- Optional post-training: If you want a model trained on your own data, that is a separate piece of work, priced per project.
The API cost crossover
Per-token API pricing is cheap to start and expensive to scale, costs rise linearly with usage and never stop. An owned on-prem system front-loads the cost, then runs at a fixed rate. Plot the two and they cross: beyond the crossover point, ownership wins and the gap widens every month.
Use the AI cost calculator to estimate your own crossover from your token volume and user count.
Locai One pricing
Locai One is a fixed-cost on-prem AI computer that includes the hardware, our open-weight Juno models and Locai OS. Locai One starts at £29,950 and Locai One Pro at £49,950, bought once and owned outright, with no per-token bills. Post-training a model on your own data is quoted separately.
What this looks like with Locai
Running AI locally means assembling hardware, a model and a serving stack that work together. Here is what it looks like when that arrives as one product.
Locai Labs builds Locai One, an on-prem AI appliance. It is one machine that arrives with everything already in it: the hardware, our open-weight Locai Juno models, and Locai OS, the operating system that serves the models and handles users, access and monitoring. You plug it into a mains socket and your network, and your team is working in about 15 minutes. No cloud account, no per-token bill, and nothing leaving the building.
The reason a data-centre-class model fits in a box on your floor is SPACE, our compression algorithm. Instead of asking how much of a model can be cut while keeping it broadly similar, SPACE asks what the model needs to be good at, preserves the subnetworks behind those capabilities and strips back the rest. The result is a smaller specialist rather than a shrunken generalist, tuned to the exact hardware it ships on.
Locai One starts at £29,950 for a team, and Locai One Pro at £49,950 for an organisation, bought once and owned outright. Both run air-cooled on standard mains power and work fully air-gapped. Any compatible open-weight model runs alongside Juno, and if you need a model trained on your own proprietary data we can post-train one and deploy it on the same machine.
Frequently asked questions
How much is an on-prem AI system?
Locai One starts at £29,950 and Locai One Pro at £49,950, bought once rather than billed per token. Post-training a model on your own proprietary data is priced separately.
Is on-prem cheaper than the ChatGPT API?
For sustained enterprise usage, yes. Per-token billing scales forever, while an owned system is fixed, so beyond the cost crossover, on-prem is substantially cheaper.
What's the payback period?
It depends on usage, but heavy enterprise workloads often reach payback within months. The AI cost calculator estimates yours.
What's included in the price?
The hardware, our open-weight Juno models, and Locai OS for serving, users, access, and monitoring, plus chat and API access.
Book a sovereign AI briefing
A 30-minute session on owning your model: deployment options, the data path, and a clear cost range for your use case.