Fuse launches AI data centre service to bypass grid delays

British energy firm Fuse has entered the artificial intelligence infrastructure market, asserting that its direct control over electricity supply enables it to construct data centres more rapidly than competitors facing protracted grid connection delays. The London based company, established by former Revolut executives Alan Chang and Charles Orr, has introduced Fuse Compute. This new service provides businesses with access to Nvidia chips hosted in UK facilities powered by Fuse’s own energy network. The initial deployment of 128 Nvidia B300 graphics processing units is scheduled to become operational in December. The company intends to expand this capacity to 1,024 units by the second quarter of 2027. Subsequent plans include adding 22 megawatts of data centre power capacity in late 2027, followed by a further 116 megawatts from 2028 onwards.

Chang stated that energy costs constitute only five per cent of GPU rental revenues, meaning that pricing is not significantly driven by energy expenses. Instead, he argued that the primary advantage lies in energy availability, which accelerates the speed at which data centres can be established. This capability allows compute customers to scale their operations more quickly. This perspective challenges the notion that cheaper renewable energy alone would provide a pricing edge in the increasingly competitive AI computing market. The move represents a significant strategic shift for the energy business, which was launched in 2022 and currently supplies electricity to 470,000 British households. The company is valued at five billion dollars, equivalent to 3.7 billion pounds.

The expansion occurs as government ambitions to position Britain as a global AI leader face increasing pressure from electricity shortages and delays in connecting data centres to the national grid. Earlier this week, Microsoft withdrew from a planned two billion pound AI data centre project in Essex. The developer, Nscale, was informed that securing sufficient grid power could be delayed until the 2030s. This setback has intensified scrutiny over Britain’s capacity to deliver the necessary infrastructure for its AI goals. Chang emphasised that Fuse has been developing energy infrastructure prior to the current AI boom, giving it a deep understanding of the sector that many other data centre developers lack. The company currently has 142 megawatts of data centre connections in its pipeline across 15 projects, alongside more than one gigawatt of planned power generation.

Fuse claims its latest solar development was constructed at a cost approximately 30 per cent below the industry average. This provides an advantage over competitors who rely on third party energy suppliers. However, these savings do not necessarily translate into cheaper AI computing services. Chang indicated that Fuse would price its GPUs competitively rather than promising to undercut established providers. The December launch will comprise just 128 Nvidia B300 GPUs, while the larger developments planned for 2027 and 2028 have yet to come online. The company has not disclosed specific pricing, noting that contracts will be negotiated individually. Fuse is also positioning itself as a British alternative to overseas cloud providers, offering domestic AI companies greater control over data processing locations. Chang described compute as critical national infrastructure, allowing British AI companies to keep sensitive data and models fully under UK jurisdiction. He added that overseas AI companies could also use Fuse’s UK facilities to serve British customers with lower latency. The company expects its near term data centre projects to represent one billion pounds of UK infrastructure investment and support 250 construction and operational jobs. This expansion follows rapid growth in its energy business, which reached group level profitability earlier this year. In June, Chang attributed infrastructure investment delays to Britain’s planning system, describing it as a self imposed bottleneck for growth. The government’s Compute Roadmap estimates that Britain will need at least six gigawatts of AI capable data centre capacity by 2030.