Civo has announced the first of 40 planned UK edge data centres as part of a nationwide infrastructure rollout targeting 1GW of capacity.
The sovereign cloud and AI provider’s first facility is in Hertfordshire, where fit-out is underway. It will open with 8MW of capacity, with scope to expand to 38MW on the same site.
Civo has also secured five further locations totalling 150MW across the UK, including sites in Lancashire, Greater Manchester, London and Oxfordshire. The remaining facilities are planned over the next three years.
The company is positioning the distributed network around AI inference, arguing that the growth of agents, robotics and other applications requiring fast responses will create demand for compute closer to where requests originate.
Civo Bets on Distributed AI Inference
Civo said the first phase of generative AI infrastructure was dominated by training, which drove demand for large, centralised compute clusters.
The cloud company said it expects inference to create a different infrastructure requirement as AI moves into applications that act and respond in real time.
“The first wave of AI was focused on training, which required enormous centralised clusters. The next wave acts,” said Civo.
Civo pointed to agents, robots, and vehicles as examples of workloads where it believes processing needs to take place closer to the requesting device. That forms the rationale for spreading its planned capacity across a national network rather than concentrating it within one large campus.
Not all inference workloads carry the same latency requirements, but the rollout represents a significant bet by Civo on more geographically distributed AI compute becoming necessary as inference demand grows.
Sovereignty Shapes Civo’s Infrastructure Strategy
The company is also positioning domestic ownership as part of the network’s appeal.
Civo has argued that distributing capacity around the UK can reduce exposure when geopolitical disruption affects infrastructure while keeping compute under British control.
“Britain has the talent. What it needs is infrastructure to match it,” said Mark Boost, CEO at Civo. “Policy alone will not deliver that. Compute, cooling and density have to exist physically, in this country, close to the organisations using them.”
Boost said the network would provide UK organisations with sovereign high-density infrastructure, while giving international organisations a UK location for services and data.
The announcement follows renewed attention on the relationship between cloud infrastructure and geopolitical risk. AWS recently said physical damage affecting multiple Availability Zones in Bahrain exceeded what its regional and multi-AZ services were designed to withstand, leaving some resources and data hosted exclusively in the affected Region inaccessible.
Civo’s rollout addresses a different infrastructure model. Domestic ownership can affect jurisdiction, control, and supply-chain dependencies, but it does not by itself guarantee resilience. Geographic redundancy, backups and recovery arrangements remain separate considerations for organisations deciding where critical workloads should run.
Sites Designed for Direct-to-chip Cooling
Civo is designing its facilities around the power and cooling requirements of newer generations of AI hardware. All 40 sites are planned for direct-to-chip liquid cooling, where coolant passes across a plate in contact with the silicon.
Civo said its in-house design team also conducts computational fluid dynamics modelling for each site before construction begins, allowing it to model how facilities should behave at full load.
The company describes its approach as scaling the module rather than continually redesigning the building as new generations of hardware arrive.
Civo said the same infrastructure will allow waste heat from its facilities to be recovered for potential use in local district heating networks serving homes, factories and greenhouses.
Civo Plans NVIDIA Vera Rubin Deployment
The facilities are being designed around NVIDIA’s Vera Rubin platform, the successor to Blackwell.
Civo said it expects to be among the first UK cloud providers to offer Vera Rubin and is already taking orders for the platform. According to the company’s announcement, early-access pricing will start at £7.50 ($10) per hour.
Civo said Vera Rubin can deliver up to five times the inference performance of Blackwell. The Vera Rubin NVL72 configuration detailed in the announcement combines 72 Rubin GPUs and 36 Vera CPUs in a single rack, with 288GB of HBM4 memory per GPU.
The company will offer Vera Rubin alongside existing NVIDIA B300, B200, H100 and H200 capacity.
The scale of the wider plan now depends on Civo progressing from its first Hertfordshire deployment and five secured locations to the full 40-site network over the next three years.
If delivered as outlined, the 1GW rollout would give Civo a sizeable distributed UK compute footprint built around its expectation that the next phase of AI infrastructure will require capacity closer to where inference workloads are being used.
Leave a comment