Fuse opens UK AI compute reservations with NVIDIA B300 GPUs

By Elena Vasquez

Fuse Energy expects the first 128 NVIDIA B300 GPUs in its new Fuse Compute service to come online at UK data centres in December, and has opened reservations. The launch moves the energy supplier into commercial AI infrastructure.

What does Fuse Compute provide?

The first version provides bare-metal, single-tenant NVIDIA GPU nodes, which Fuse hosts in UK data centres. The company will negotiate pricing, terms and delivery under individual contracts, suggesting that the opening phase targets buyers seeking dedicated capacity rather than shared, off-the-shelf cloud access.

The service will operate under UK jurisdiction and use a UK data path. For organisations with data-location and legal-control requirements, that jurisdiction and routing may determine whether the service meets those conditions.

• December: Fuse expects its first 128 NVIDIA B300 GPUs online.
• Second quarter of the following year: The rollout scales to 1,024 GPUs from the initial 128.
• Fourth quarter: A further 22MW follows.
• From 2028: Fuse plans 116MW.

Why does power access matter?

The launch rests on Fuse's view that power access is the binding constraint. Alan Chang, Chief Executive Officer and Co-Founder of Fuse Energy, said: “Power is the biggest hurdle to building AI infrastructure, and it's the part we already know how to do.”

Fuse's approach combines its energy operations with data-centre access. The company argues that controlling energy assets and grid connections can shorten the time required to bring compute online. It supplies 470,000 UK homes and holds both supplier and meter-operator licences.

Across its business, Fuse employs more than 600 people in energy generation, supply and software. Alan Chang and Charles Orr, both Revolut alumni, founded the company. Fuse says its AI strategy covers the full chain from power and land to data centres and compute.

Fuse reports a power-generation pipeline of more than 1GW and a data-centre connection pipeline of 142MW across 15 projects. Its wider generation portfolio includes 22MW of active capacity and more than 1GW in the pipeline across the UK, US and Europe.

Fuse says its most recent solar site was built at around 30% below the industry-average cost per megawatt. It presents that figure as evidence that it can deliver energy infrastructure at lower cost than some rivals.

Fuse expects its near-term data-centre projects to bring 250 construction and operational jobs to the UK. The company says those projects represent £1 billion in UK infrastructure investment.

How does the grid constraint change deployment?

The UK government estimates that Britain needs at least 6GW of AI-capable data-centre capacity by 2030. Fuse says waits for grid connections at new data centres can stretch to a decade.

As AI compute demand has risen sharply, long delays in securing connections have constrained new data-centre development. Land, electricity and grid connectivity now play a major role in deciding which projects can proceed quickly, alongside procurement of advanced NVIDIA processors. Start-ups, research groups and larger companies building and training models all compete for that supply.

Chang also addressed GPU scarcity. He said: “Every AI company is fighting for the same scarce GPUs, and the grid can't connect data centres fast enough to keep up. Fuse Compute puts that expertise to work for the companies building the future of intelligence, and makes sure that future is built in Britain.”

What does domestic capacity change?

Britain is trying to expand public compute resources, but domestic supply remains limited, leaving many UK AI businesses dependent on infrastructure hosted abroad. Policymakers and technology companies have raised resilience, sovereignty and whether British AI firms can scale locally as recurring concerns. Fuse's first release gives buyers a domestic option with UK hosting, jurisdiction and a UK data path.

The launch also moves Fuse beyond consumer and energy operations into AI hardware access, where tighter chip supply has met surging demand for advanced processors. Buyers must therefore assess processor access alongside the power and grid schedule behind each deployment.