SWI Compute
Category: AI Infrastructure
SWI Compute is the AI cloud and compute subsidiary of Euronext Amsterdam-listed SWI Group. It builds and operates NVIDIA AI factories, starting in Europe. SWI Compute was founded in 2026. The company is led by Halil Mirakhmedov. Latest round: Undisclosed. Key investors include SWI Group (parent, Euronext Amsterdam: SWICH).
- Founded
- 2026
Value proposition
SWI Compute converts the parent group's 3.6 GW of secured power, spread across 24 sites in Europe and the US, into dedicated and on-demand NVIDIA GPU capacity. It sells that capacity to AI natives, AI labs, universities and enterprises under European jurisdiction.
Products and solutions
NVIDIA AI factories offered as dedicated and on-demand GPU cloud (GPU-as-a-Service) for training, inference, fine-tuning and agentic AI workloads. The first deployment runs NVIDIA Blackwell Ultra systems, with more than 8,000 GPUs integrated by Delta Computers. The first UK factory starts at 30 MW with about 7,000 Rubin GPUs and 3,500 Vera CPUs in Vera Rubin NVL72 racks, and is due online by end of 2027. New sites from 2027 onward will follow the NVIDIA Vera Rubin DSX AI Factory reference design.
Unique value
A vertically integrated stack from grid connection to GPU, built on power and land that SWI Group spent six years securing. It offers sovereign European compute built to NVIDIA reference architectures.
Target customer
European AI startups and AI-native companies, AI labs, universities, enterprises, and AI inference platforms. Its first contracted customer is a US-based AI inference platform.
Industries served
AI cloud / GPU compute, AI labs, enterprise AI, AI inference platforms, research and academia
Technology advantage
Access to the parent group's 3.6 GW of power capacity, including AiOnX's five European campuses (about 2 GW) and SWI Digital's US sites. Preferred NVIDIA Cloud Partner since August 2026. About $1B of NVIDIA systems ordered within two months. Adopting the NVIDIA DSX full-stack AI factory platform.
How they differentiate
Unlike neoclouds that lease data center capacity, SWI Compute gets power and land through its sister companies. This gives it faster time-to-power, with 2.3 GW of capacity planned by 2029 and a 530 MW UK campus. It also emphasizes European data residency and jurisdiction.
Main competitors
Nebius, Nscale, Fluidstack
Key partnerships
NVIDIA (Preferred NVIDIA Cloud Partner, DSX AI Factory reference design), Delta Computers (systems integrator), AiOnX (sister European data center platform), SWI Digital (sister US data center platform), Polarise (receives financing from SWI Group)
Notable customers
Unnamed US-based AI inference platform (first contracted customer)
Major milestones
Aug 2026: became a Preferred NVIDIA Cloud Partner. Oct 2026: emerged from stealth at NVIDIA GTC Berlin and announced the ~$1B NVIDIA systems order. End of 2027: target for the first UK AI factory to come online. 2029: target of 2.3 GW capacity.
Growth metrics
About $1B of NVIDIA hardware orders. Planned AI factory capacity of 2.3 GW by 2029. The first UK site starts at 30 MW and is planned to grow to 530 MW. The first deployment has more than 8,000 GPUs. The parent group holds 3.6 GW of power across 24 sites.
Market positioning
A newly launched European neocloud and AI factory operator. It sits inside a listed, vertically integrated digital infrastructure group and targets demand for sovereign European AI compute.
Geographic focus
Europe first (UK, Ireland, Denmark, Spain, Italy), with global deployment ambitions including the US.
About Halil Mirakhmedov
Co-founder and director of the Mirakhmedov Foundation, which he founded with his father Abdumalik Mirakhmedov (founder of SWI Digital). Confirmed as CEO of SWI Compute in its October 2026 launch press release and coverage dated Oct 26, 2026.
Latest news about SWI Compute
More AI Infrastructure companies
Official website: https://swi.com/our-strategy/