Singapore and Amsterdam, October 26th / PRNewswire / -- SWI Compute has officially come out of its "invisible" state today, beginning the construction and operation of the NVIDIA AI factory, with the first site located in Europe. Behind this global AI infrastructure company is a power capacity of 3.6 gigawatts, which SWI Group took six years to secure in Europe and the United States. SWI Compute is a new company under the umbrella of the publicly listed Euronext Amsterdam corporation SWI Group; it became a NVIDIA Preferred Cloud Partner in August and has already placed orders for NVIDIA systems worth approximately 1 billion US dollars, signing final contracts with its first batch of customers. SWI Compute will be deployed globally, but its presence and implementation in Europe are particularly strong.
Its first large-scale AI factory in the UK started with a capacity of 30 megawatts, equipped with nearly 7,000 Rubin GPU and 3,500 Vera CPU. It uses NVIDIA Vera Rubin NVL72 racks and is expected to be put into operation by the end of 2027. These 30 megawatts are connected by the NVIDIA Quantum - X800 InfiniBand and NVIDIA Spectrum - X Ethernet networks and are part of the first phase of the project in this UK park; the park is planned to be expanded to 530 megawatts, with the goal of achieving a factory capacity of 2.3 gigawatts by 2029.
Energy availability is the primary bottleneck facing European AI infrastructure. It took SWI Group six years to obtain grid access, and now they are deploying these power resources into dedicated NVIDIA AI factories through SWI Compute. Just two months after joining the NVIDIA Cloud Partner initiative, our team received orders for advanced hardware worth approximately one billion US dollars and signed the first customer contracts. Market demand consistently exceeds available production capacity, so we are rapidly expanding our operations to meet this demand," said SWI Compute, the CEO of Halil Mirakhmedov.
Europe has researchers, founders, and enterprises capable of creating cutting-edge AI. However, a large portion of the infrastructure development has taken place outside of Europe, and the existing capacity is primarily occupied by global hyperscale cloud service providers and leading laboratories, which are able to sign large-scale multi-year contracts. The companies that are most in need of computing power—AI native enterprises, as well as startups with ambitious goals but limited financial resources—are the ones that have the greatest difficulty in obtaining these resources.
SWI Compute is built precisely for these customers. It provides exclusive and on-demand access to NVIDIA AI factories for native European AI enterprises, AI laboratories, and enterprise clients. By being located close to users and data and under European judicial jurisdiction, it enables the next generation of AI to be trained, fine-tuned, and deployed in its place of origin. Its first customer is a AI inference platform headquartered in the United States, for which capacity has already been contracted.
SWI Group has been investing in land, power, and data center infrastructure since 2020. Today, the group owns 3.6 gigawatts of power capacity in Europe and the United States, with a particularly strong presence in Europe, where it has five data centers located in Ireland, the UK, Denmark, Spain, and Italy. One of these data centers has been leased to a large hyperscale cloud service provider. The group develops projects from the point of connecting to the power grid to the actual deployment of the facilities. Power supply agreements have been signed, and the systems have already been ordered.
Each site of SWI Compute is constructed in accordance with the NVIDIA reference architecture. In the future, SWI Compute will adopt the complete NVIDIA DSX AI Factory Platform to design, construct, and operate its AI factories, where computing, networking, power supply, and cooling systems will be co-designed as an integrated whole. Starting from 2027, all new sites will follow the NVIDIA Vera Rubin DSX AI Factory reference design, beginning with the sites in the UK. Based on NVIDIA Vera Rubin NVL72, DSX will be able to achieve up to a 40% increase in GPU capacity within the same power budget.
The first batch of deployments will use the NVIDIA Blackwell Ultra system, followed by the introduction of NVIDIA Vera CPU racks, which will be connected through NVIDIA Quantum - X800 InfiniBand and NVIDIA Spectrum - X Ethernet. SWI Compute has selected system partners from the ecosystems of NVIDIA's OEM and ODM, including Delta Computers as the system integrator, as well as well-known storage partners in the industry, to deliver over 8,000 GPU.
The revenue of the AI factory depends on three factors: how much token can be produced per megawatt, how many years it can continue to produce, and how many types of tasks it can handle. Over time, continuous software optimization ensures that the installed system can continue to generate revenue even years after delivery. Across different workloads, the same infrastructure can run all types of tasks at various stages of AI, from pre-training to agent inference, and can also be used for simulation, data processing, and scientific computing. For customers of SWI Compute, this means that the computing power can remain up-to-date; for SWI Group, it means that the assets can continue to be productive.
"The AI factory is a critical infrastructure that can transform regional energy into intelligent and economic opportunities," said N, the Vice President of Global AI Infrastructure Growth at NVIDIA. "We are putting Europe's energy into the ecosystems of native enterprises, laboratories, universities, and businesses through SWI Compute. Based on the NVIDIA reference architecture and adopting NVIDIA DSX in the next generation of AI factories, we are building an infrastructure that is productive, durable, and replaceable – one that can serve new customers and workloads, and continue to create value for many years to come."
The first AI factory is just the beginning of SWI Compute's ambitious expansion strategy. The company plans to start rolling out its next-generation AI facilities in Europe from the fourth quarter of 2027, deploying thousands of GPU.
The competitive advantage of SWI Compute lies in the fact that it is built upon the power infrastructure already controlled by SWI Group, allowing for the connection of up to 24 sites with a total power capacity of 3.6 gigawatts.
This enables SWI Compute to expand rapidly and transform the group's existing energy infrastructure into the next generation of AI cloud platform.
"We started investing in land, power, and data center infrastructure in 2020. SWI Compute is the last piece of the puzzle, providing a fully integrated infrastructure for the NVIDIA AI factory. SWI Group is now a transatlantic digital infrastructure company with vertical integration, extending from grid access all the way to GPU, while SWI Compute delivers this capacity directly to the customers who need it most," said SWI Group founder and CEO Max-Herv é George.
About SWI Compute
SWI Compute is a AI with its headquarters in Europe, and it serves as an infrastructure provider, belonging to SWI Group ( Euronext Amsterdam : SWICH ). SWI Compute provides computing power services through its data centers in AiOnX in Europe and SWI Digital in the United States, complementing SWI Group's transatlantic, vertically integrated AI technology stack. For more information, please visit www.swi-compute.com and www.aionx.com.
About SWI Group
SWI Group ( SWI Capital Holding Ltd ) is an investment group that spans the Atlantic Ocean, allocating over 80% of its capital to digital infrastructure. It is listed on the Euronext Amsterdam with the stock code SWICH. The group was formed by the merger of token issuers na and Stoneweg, and it deploys its own capital primarily in digital infrastructure, real estate, and other private market opportunities, combining entrepreneurial approaches with institutional discipline. For more information, please visit www.swi.com.
