Summary
- Fuse has opened reservations for single-tenant GPU infrastructure across 15 proposed UK projects representing approximately 142MW.
- The company's published timetable starts with 0.7MW in December 2026 and includes larger phases during 2027 and from 2028.
- The pipeline represents intended development, not 142MW of commissioned computing capacity; individual connection arrangements and delivery milestones remain undisclosed.
British energy company Fuse has launched a commercial GPU computing offering, opening reservations for infrastructure it intends to deploy across 15 UK data centre projects with approximately 142MW of combined planned capacity.
The service, called Fuse Compute, will provide single-tenant NVIDIA GPU systems hosted in Britain. The company is advertising its first availability for December 2026, followed by larger deployment phases during 2027 and from 2028 onwards.
Fuse Energy is extending its existing electricity generation and supply activities into the physical infrastructure required to host AI computing. Its commercial proposal brings electricity expertise, data centre development, and dedicated computing hardware under one business model.
Fuse’s published capacity schedule begins with a 0.7MW allocation targeted for December. A second phase of 3.25MW is advertised for the second quarter of 2027, followed by 22MW in the fourth quarter and a further 116MW from 2028.
Together, those allocations total 141.95MW, rounded by the company to a 142MW development pipeline. The timetable is indicative, and Fuse states that availability, pricing, terms, and delivery must be agreed through individual customer contracts.
The company identifies NVIDIA B300 systems for its first two phases. Hardware selection for the larger later allocations has not been specified and will depend on contractual arrangements made closer to delivery.
The reservations represent an invitation to secure future computing services rather than evidence that the entire infrastructure programme has been constructed or commissioned. Fuse has not published a project-by-project list of completed data centres or disclosed the operating status of every site within its pipeline.
Its new business will provide bare metal computing, allowing customers to use dedicated GPU hardware rather than relying exclusively on virtualised shared cloud resources. The offering is intended for organisations requiring control over computing configurations and the ability to run their own software environments.
The planned infrastructure is described as sovereign UK compute. Fuse says the installations will be hosted in Britain, operated by a British company, and designed to keep customer data within UK jurisdiction.
Those characteristics may help businesses meet certain contractual, security, and data location requirements. However, physical location alone does not establish complete technological independence, since computing services also depend on hardware suppliers, software, network providers, security arrangements, and operating controls.
Fuse’s proposition is also closely linked to the availability of electricity. Large computing installations require dependable electrical connections and sufficient capacity for the IT equipment, cooling plant, power conversion, and supporting infrastructure.
The company argues that its experience developing electricity generation and managing energy supply gives it capabilities that conventional computing developers may need to obtain from external partners.
Fuse has described a broader low carbon generation pipeline exceeding 1GW, alongside its proposed data centre connections. These figures concern different infrastructure programmes and cannot be treated as interchangeable capacity.
Owning or developing renewable generation does not automatically give a data centre an operational electricity connection. Power must reach the site through appropriate network infrastructure, approved private arrangements, or another technically and commercially viable supply configuration.
The company has not disclosed how each of the 15 proposed data centre projects will be supplied, which grid connection dates are secured, or whether individual installations will combine utility connections with private generation or storage.
That distinction is important for understanding the 142MW figure. Planned electrical capacity, reserved grid capacity, installed IT load, and live computing output represent different stages of development and cannot be added together without identifying their definitions.
Even after a suitable power supply is available, GPU deployments require appropriate electrical distribution, cooling, network connections, fire protection, monitoring, and physical security. Those systems must be commissioned alongside the computing hardware before customers can use the facility.
High-density accelerators can place substantial demands on rack power delivery and thermal management. The resulting heat must be removed through cooling systems capable of operating under the anticipated workload and local environmental conditions.
Fuse has not published the detailed cooling designs, rack densities, measured efficiency figures, or redundancy arrangements for its proposed computing installations. The advertised power allocation therefore does not establish the number of operating racks or the facility’s total electricity consumption.
The company has also outlined a substantial investment programme for its new activity. Reporting published on 21 October indicates an ambition to invest approximately £1 billion in UK data centre infrastructure during 2027, although a detailed allocation of committed capital between projects has not been disclosed.
The investment ambition reflects the cost of building and equipping computing facilities rather than solely purchasing GPUs. Land, electricity connections, technical buildings, power and cooling plant, network infrastructure, and specialist engineering services all contribute to delivery.
Fuse’s proposed expansion comes as other British computing developments encounter significant delays securing electricity connections. Microsoft’s withdrawal from a planned Nscale installation in Essex has illustrated how a commercial computing agreement can be affected when grid availability falls behind the proposed delivery date.
Fuse’s individual projects will face their own engineering and commercial requirements, and the company has not established that its electricity capabilities remove those constraints across the entire portfolio.
The first advertised deployment is much smaller than the full programme. Its 0.7MW December allocation represents approximately half of one per cent of the 142MW pipeline, leaving the overwhelming majority of planned capacity for later phases.
Delivery of the initial installation will provide a more immediate measure of the company’s move into commercial computing. Subsequent milestones will include confirmed customer agreements, energised facilities, installed hardware, operating service availability, and construction progress on the larger projects.
Fuse has established a new commercial offering and published an indicative capacity programme. The scale of the business will become clearer as the company converts its proposed electricity and data centre infrastructure into commissioned computing systems serving paying customers.
