Training and fine-tuning
Multi-GPU nodes with high-bandwidth interconnect for training runs and fine-tuning jobs. Reserve by the hour, scale up for a run, hand the capacity back when it finishes.
The UK's distributed AI cloud
Meridian is building a UK network of compact nodes running enterprise NVIDIA accelerators, sited where the electrical capacity already exists. Each node is designed to come online in days rather than years, and to keep your data in the country.
We design, deploy and operate the compute that runs the next generation of British AI.
AI cloud platform
Multi-GPU nodes with high-bandwidth interconnect for training runs and fine-tuning jobs. Reserve by the hour, scale up for a run, hand the capacity back when it finishes.
Serving from nodes distributed across the UK rather than a single region, so round-trip latency is measured against where your users actually are.
Deploy an open-weight model behind an endpoint without provisioning the machine underneath it. Orchestration, scheduling and failover are handled by us.
Meridian's activity sits within SIC 63110 — data processing, hosting and related activities. We supply compute capacity; we do not sell credits, tokens or hardware.
63110
Our sole SIC code
100%
UK data residency
£0
Egress charge on your own data
The Network
A Meridian node is a compact, liquid-cooled rack installed at a site that already has the electrical capacity to run it. We install and own the equipment and operate it remotely, so nothing about running it falls to the host.
Because each site is small, it can be commissioned in weeks rather than years. Because there are many of them, capacity lands near the users who need it. Every node carries battery backup so work rides through an interruption instead of failing.
Installed where spare electrical capacity already exists, so there is no grid connection to queue for and no new supply to build.
Closed-loop liquid cooling, quiet in normal surroundings and far denser per square metre than air.
Remote orchestration, monitoring and failover are designed in from the start, so a job reschedules to another node rather than failing.
You submit to a single UK endpoint. The scheduler places the job on the node with free capacity nearest your users, streams logs back, and releases the GPU the moment the job ends — so you are billed for the run, not the reservation.
What we are, plainly
Meridian is not an energy supplier or generator. We do not generate, supply, trade or resell electricity. We buy power to run our own equipment, the same as any other business that owns computers.
We are not a property developer and not a bitcoin or digital-asset miner. Our revenue comes from one thing: renting GPU compute to businesses by the hour, invoiced in sterling.
About the companyBuilt for the demands of AI
Early access is open to UK businesses running training or inference workloads. Tell us what you need and we will give you a straight answer on what we can serve, and when.