Built for
Sovereign AI.
A real computer, engineered to run your own models on GPUs inside your walls. No cloud, no per-token bills, no data leaving the building.
And it's fully managed, the models, the application suite and the support are all handled for you, so your team just uses it.
The on-prem sovereign AI solution for enterprise and government.
The problem, and why now
Renting AI means
renting the risk too.
Your intelligence lives on someone else's computer.
Every prompt, document and customer record is processed in someone else's data centre, under foreign jurisdiction and laws you don't control. Sovereignty is now national policy, and procurement is following.
Your data leaves the building to get an answer.
Confidential and regulated data is sent to a third party on every request. For finance, healthcare, legal and government, boards and regulators now treat that cross-border exposure as a live risk.
You rent intelligence, and the bill never stops.
Per-token pricing scales with every user and workflow. As AI spreads across your teams it compounds every quarter, and you own nothing at the end of it.
The model can change or vanish underneath you.
Vendors deprecate models, raise prices and change terms on their schedule, not yours. The capability your business runs on isn't actually yours.
Locai One solves all four at once, by putting the model, the data and the compute inside your own walls.
Frontier AI.
In one computer.
Run Locai Models or any open source model.
GPU: 1–4 enterprise GPUs
Up to 96GB VRAM each. Inference on-prem.
CPU: Server-class CPU
High core count for host compute.
RAM: High-bandwidth DDR5
Dual-channel and expandable.
Storage: Up to 8TB NVMe SSD
Configurable capacity. Data stays local.
Power: Efficient, redundant power
Clean, quiet power with headroom for a second GPU.
Network: Multi-gig networking
Serves your whole team across your local network.
Choose your
power.
concurrent users
Built for teams getting started with sovereign AI on-prem.
- Single enterprise GPU
- 96GB GPU memory
- Desk-side workstation form factor
- Serves a full team over your LAN
concurrent users
More headroom for larger models and heavier concurrent load.
- Dual enterprise GPUs
- 192GB GPU memory
- Heavier models and higher concurrency
- Scale with additional units as you grow
concurrent users
For the whole organisation, scaled across multiple units.
- Multiple GPU units
- Run several 300B-class models
- Thousands of concurrent users
- Scale out as demand grows
Everything you need.
In one place.
Not just hardware, a fully managed sovereign-AI service. The computer, the models, the application suite and the support, all handled for you.
One platform · every modality
Draft, summarise, answer questions and analyse documents. The everyday assistant for your whole team.
A fully managed service
We run it, end to end.
Deployment, tuning, monitoring and model operations are handled by our team. No GPU wrangling, no drivers, no MLOps hire required.
Ready for your team on day one.
A chat interface, an OpenAI-compatible API, and the admin and usage platform come built in, so the whole organisation can start working immediately.
The right models, kept current.
We select, install and update the best models for each capability, your sovereign Locai model plus a curated library, so you're never left maintaining them.
Backed by Locai, continuously.
Ongoing support, security updates and model refreshes, with a team that knows your deployment, for the life of the unit.
Getting started
Plugging in to using
your AI in 10 minutes.
On-prem doesn't have to be a project. Setup takes about ten minutes, plug it in and you're good to go, or we'll come to you and do it.
Order & configure
We confirm your configuration and what your team needs, then start your build.
Built, tested & delivered
Your Locai One is assembled, loaded with your models and the application suite, and shipped to you.
Plug it in
Connect power and network, that's it. Setup takes about ten minutes and you're good to go.
Or we come to you
Prefer a hand? Our team can come to your site and set everything up for you.
Locai OS
Your expert AI,
on every surface.
Locai OS is the software stack that ships on every Locai One. Run Locai's own models or any open-source model, and serve them to your whole organisation, all inside your perimeter, with nothing leaving your network. Reach it from an API, a chat UI or the control panel, whichever fits the workflow.
API
An OpenAI-compatible REST API that serves every employee fast, frontier-class AI. Drop it into existing tools, apps and services without rewriting integrations.
Why own intelligence?
The more you use it, the cheaper it gets
Unlike per-token cloud AI pricing, the cost does not rise every time your team uses it. Once the unit is paid for, usage can scale internally without every prompt becoming another bill, so the more your organisation runs on it, the lower your effective cost per output drops.
Nothing leaves the building
Locai One runs entirely inside your environment, on your hardware or in your own cloud account. Your data, your prompts and your model weights never leave your network.
Run any model you want
Locai's own frontier models or any open-source model: Llama, Qwen, DeepSeek and more. Swap, upgrade and run whatever your teams need, all on hardware you own.
Your whole team, served locally
Custom inference and concurrency engineering keeps everyone fast, so the model responds in real time even when the whole team is asking at once.
Configure Your Locai Computer.
GPUs. Choose from these powerful options.
Storage. Pick how much is right for you.
Model. Choose what runs on your Locai Computer.
Locai OS. Included free, limited launch offer.
Locai OS included
API, chat UI, the usage platform and the full application layer, ready to serve your team.
- Product
- Locai One
- GPUs
- 1 GPU
- Storage
- 4 TB
- Model
- Locai & OSS
- Software
- Locai OS · Included
We'll put together a fixed-cost quote for this configuration, sized to your team.
Locai One vs cloud AI APIs
Owned, not rented.
How a Locai One you own compares to renting intelligence from OpenAI, Anthropic, Azure OpenAI and the rest.
How much can you
save with Locai One?
Every month, you rent intelligence from someone else's cloud. Own it once and keep the difference, year after year.
Regular agent workflows, document analysis and long-context work. Pro-tier AI tools plus meaningful API spend per user.
iWhat we assume
- Per-user figures represent monthly token and API spend for popular cloud assistants and agent tools.
- Each Locai One serves up to 250 users, so larger teams run a small fleet. The unit is a fixed asset you own, so it isn't counted against your savings.
- 3-year saving = three years of cloud spend you no longer pay. Per-member saving divides that across your team.
- Your real numbers will vary by team, workload and contract. We'll model your exact picture in a quote.
Saved over 3 years
The cloud spend you stop paying once your team runs on a Locai One you own outright, a fixed asset, with no per-token bills.
Saved per team member
Across 50 users · over 3 years
Annual cloud spend avoided
50 users · £200/user/mo
Three-year cloud spend avoided
Cumulative · the meter stops running
Indicative figures. Current-tools spend is modelled as monthly token and API consumption per user; your real numbers will vary by team and workload.
Where Locai One
is at home.
- 01
Financial Services
Trading, ledger and risk workflows with full audit lineage on every AI-assisted decision.
Explore - 02
Public Sector
Government teams that need intelligence to stay inside the building.
Explore - 03
Healthcare
EHR integrations, clinical apps and protected algorithms that need patient data to stay on your own infrastructure.
Explore - 04
Energy & Utilities
Operational tech, grid systems and critical-infra knowledge that can't ship to a public API.
Explore - 05
Manufacturing
ICS, production systems and operational knowledge that stays sovereign by hardware, not just by policy.
Explore - 06
Telecommunications
Core network operations, BSS/OSS and customer-data workflows kept inside your perimeter.
Explore - 07
Science & Research
Pre-publication research, experimental models and IP that can't leak to third-party clouds.
Explore - 08
Technology
Enterprise AI for organisations that won't ship their data to a vendor.
Explore
Own the model.
Own the hardware.
Own the intelligence.
Tell us your team size, your use cases, and your compliance constraints. We'll model your ROI, set up the models your team needs, and have a working Locai One on your floor in weeks.