Summary
- The new UK national supercomputer will be hosted by the University of Edinburgh and owned by UKRI.
- The system is expected to be around 50 times more powerful than ARCHER2.
- The facility design includes heat reuse for university buildings and potential local district heating.
EPCC has marked the start of construction on the buildings that will house the UK’s new £750m national supercomputer near Edinburgh.
The system will be owned by UK Research and Innovation and hosted by the University of Edinburgh at a site near Penicuik and Roslin in Midlothian. It is expected to be around 50 times more powerful than ARCHER2, the current UK national supercomputing service.
The facility is expected to support advanced workloads in areas including weather modelling, drug discovery, aircraft engineering, and scientific simulation. Its delivery now moves from policy commitment into the physical build: buildings, power systems, cooling, heat-reuse connections, network infrastructure, and security.
Sovereign compute needs plant rooms
National supercomputing is often framed around performance, research capacity, and scientific use cases. The Edinburgh project also has to be built as a critical facility, with enough electrical and mechanical resilience to operate dense compute at sustained load.
The location gives the project room for specialist infrastructure outside the city centre while keeping it close to the University of Edinburgh’s established HPC expertise. It also brings familiar questions for advanced compute: where the power comes from, how cooling is designed, how heat is rejected, and how future hardware refreshes will be accommodated.
Surplus heat is expected to be used in university buildings, with assessment also under way for potential district heating to local homes. That adds a useful infrastructure layer, but heat reuse depends on temperature, demand profile, distance, network economics, heat pumps, and seasonal operation. A stated heat-use route has to become engineered pipework and commercial agreements.
The UK compute race becomes a build programme
The UK government has been under pressure to expand sovereign compute as AI, climate modelling, defence, health research, and advanced science place heavier demand on national infrastructure. Supercomputers and AI research clusters are now part of industrial strategy as well as research funding.
Delivery depends on more than procuring processors. High-performance computing facilities need specialist contractors, secure supply chains, long-lead equipment, commissioning expertise, and operational teams able to manage dense, power-hungry systems. The machine can only perform if the building services keep pace.
Cooling will be a defining part of the facility. Dense HPC systems increasingly move beyond conventional air-cooling assumptions, requiring high-capacity liquid or hybrid thermal approaches. Public information so far has focused on heat reuse rather than the full mechanical specification, but the final design will have to balance performance, resilience, and efficiency.
The Edinburgh build also sits inside a wider European race for AI and HPC capacity. Public-sector supercomputing has different governance and workload priorities from commercial GPU cloud, yet it competes for many of the same physical inputs: power, cooling plant, skilled labour, and high-value equipment.
With construction now under way, the next markers will be procurement, supplier selection, cooling detail, commissioning schedule, and evidence that heat reuse can become an operating asset rather than a side claim. The UK’s sovereign compute strategy now has a construction programme attached to it.
