SWI Compute Plans 30MW NVIDIA AI Factory in UK

A new AI cloud company backed by the Amsterdam-listed SWI Group says it will open its first large AI factory in the UK by the end of 2027, after ordering about $1bn of NVIDIA systems.

SWI Compute came out of stealth on 26 October. In its launch announcement, the company says the UK site will start at 30 megawatts (MW) with close to 7,000 NVIDIA Rubin GPUs and 3,500 Vera CPUs in Vera Rubin NVL72 racks. SWI Compute says that is the first phase of a UK campus planned to grow to 530MW. It has not said where the campus is.

In brief

SWI Compute says it has ordered about $1bn of NVIDIA systems and signed its first customer, an unnamed US AI inference platform.
Its first big AI factory, in the UK, is due online by the end of 2027 at 30MW, with a plan to grow the campus to 530MW.
The company has not named the UK site or said how the hardware is being paid for.

SWI says power, not chips, is its advantage

SWI Compute’s pitch rests on electricity. Its parent, SWI Group, says it began buying land, power and building shells in 2020 and now holds 3.6 gigawatts (GW) of power capacity across 24 sites in Europe and the US. In Europe it has five campuses, in Ireland, the UK, Denmark, Spain and Italy, and says one of them is already leased to a major hyperscaler, which it does not name.

“Energy availability represents the primary bottleneck facing European AI infrastructure,” said Halil Mirakhmedov, chief executive of SWI Compute. He said the team secured about $1bn in hardware orders and signed initial client contracts “just two months after joining the NVIDIA Cloud Partner program”. The company became a Preferred NVIDIA Cloud Partner in August.

SWI Compute says it is aimed at what it calls Europe’s “AI natives”: younger AI companies and labs that struggle to win capacity because data centre space has gone first to hyperscalers and frontier labs able to sign long contracts. Its stated target is 2.3GW of NVIDIA AI factory capacity by 2029, with a next wave of European campuses coming online from the fourth quarter of 2027.

The first systems are Blackwell Ultra, with DSX from 2027

The company’s first deployment runs NVIDIA Blackwell Ultra systems, with Delta Computers as systems integrator, and will total more than 8,000 GPUs, according to the release. Every new site from 2027, starting with the UK campus, will follow NVIDIA’s Vera Rubin DSX AI Factory reference design. SWI quotes NVIDIA’s claim that DSX can fit “up to 40% more GPU capacity within the same power budget”. TSN covered NVIDIA’s push to sell whole AI factories as repeatable units last week.

“SWI Compute is putting Europe’s energy to work for the ecosystem of AI natives, labs, universities and enterprises,” said Nico Caprez, NVIDIA’s vice president for global AI infrastructure growth.

SWI Group is listed on Euronext Amsterdam under the ticker SWICH and was formed from the merger of Icona and Stoneweg. Founder and chief executive Max-Hervé George said the group is now “a transatlantic, vertically integrated digital infrastructure company, from the grid connection to the GPU”.

A 530MW UK campus would be a big claim on the grid

For the UK, the headline numbers are large. The planned 530MW campus is almost 18 times the size of the 30MW first phase, and getting a grid connection is the step SWI Compute’s own chief executive calls the main bottleneck. That makes the missing location the most important gap in the announcement.

SWI’s own campus business, AiOnX, does list a UK site. Its website shows a “Cambridge Innovation Campus” just north of Cambridge in pre-development, saying it has “already secured 330 MW from the grid”. That is 200MW short of the 530MW the release describes, so either the UK campus is somewhere else, or the plan has grown beyond what AiOnX has published. SWI Compute has not said which.

The money is the other open question. About $1bn of NVIDIA hardware is a large order for a company launched the same day, and the release says nothing about how it will be paid for. SWI Group did announce an €800m share sale to institutional investors on 12 October, which it said would fund data centre and AI compute growth. Rivals have reached for debt: Nebius took a $3.5bn bridge loan from Citi to pay for its GPU build-out. Big cloud providers are racing for the same chips, with Google Cloud setting a launch date for its Vera Rubin service.

The tests are straightforward: a named UK site with a confirmed grid connection, a planning application, and a first customer willing to be identified. None has been made public yet.