Civo Unveils 150 MW Distributed AI Edge Network Across the UK
Cloud provider begins 40-site roll-out with 8 MW Hertfordshire facility fitted for liquid-cooled Nvidia Vera Rubin racks
UK cloud vendor Civo has announced a major expansion of sovereign AI compute infrastructure, starting with an 8 MW edge data center in Hertfordshire. The facility marks the first of 40 planned edge sites across the country designed to provide distributed high-density computing and bypass centralized power grid bottlenecks.
Key details
The Hertfordshire facility is currently being fitted with direct-to-chip liquid cooling to support high-density hardware, including Nvidia Vera Rubin NVL72 architectures. Operating at an initial 8 MW capacity, the site is designed to scale up to 38 MW as demand grows.
In addition to Hertfordshire, Civo has secured five additional UK locations across Lancashire, Greater Manchester, London, and Oxfordshire, which together represent 150 MW of planned compute capacity. Over the next three years, the company intends to complete a distributed network of 40 edge data centers by leveraging brownfield and industrial sites in partnership with infrastructure firm Era4.
Why this matters
As hyperscale data centers face severe grid queue delays and growing public opposition in major hubs like London and Slough, distributed edge architectures offer an alternative path for AI deployment. By decentralizing power demand across 40 smaller sites, Civo aims to deliver high-density sovereign AI infrastructure without requiring massive single-site grid upgrades or localized water infrastructure draw.
Context
The UK energy grid is currently struggling with a backlog of data center connection requests exceeding 73 GW, leading regulator Ofgem to propose stricter financial criteria to clear "phantom" applications. Regional edge deployments that utilize existing brownfield industrial power hookups allow AI cloud providers to rapidly scale infrastructure while maintaining compliance with domestic data sovereignty requirements.
Risks and open questions
Deploying 40 distributed edge facilities poses operational and economic challenges, particularly around maintaining high power usage effectiveness (PUE) across smaller sites. It remains uncertain whether edge sites can secure long-term power purchase agreements at competitive prices compared to mega-scale facilities, or if localized power grids in regional areas will accept high-density liquid-cooled workloads without requiring additional infrastructure surcharges.
What happens next
Civo will continue fitting out the Hertfordshire facility and expects to begin onboarding sovereign AI workloads later this year. Development across the five secured regional sites in Lancashire, Greater Manchester, London, and Oxfordshire will proceed as the company finalizes site acquisitions for the remaining 34 planned edge locations over the next 36 months.
Source: Data Center Dynamics Published on AI Usage Global, author: AUG Bot
