Data center developer AiOnX closed a $500 million deal on Monday to acquire a 77% stake in Genesis Digital Assets, a US-based Bitcoin mining operator, and will convert all 15 of its facilities to serve AI and high-performance computing workloads — one of the largest crypto-to-AI infrastructure acquisitions completed in 2026. The deal was confirmed by SWI Group, AiOnX's Euronext Amsterdam-listed parent company, which manages roughly €11 billion ($12.7 billion) in assets under management.

The timing reflects a structural supply problem. AI data center capacity in both the United States and Europe has grown acutely scarce as hyperscalers and GPU cloud providers race to secure sites with reliable grid connections. Of roughly 140 large-scale US data center projects representing approximately 12 gigawatts of planned capacity scheduled to go live in 2026, only about a third are currently under construction. Mining operators, who spent the 2021–2024 Bitcoin buildout acquiring powered, connected land at scale, now hold what the AI industry most urgently needs: energized substations, approved grid connections, and operational physical infrastructure — the three items that bottleneck new construction.

Genesis Digital Assets: Fifteen Sites, One Point Three Gigawatts

Genesis Digital Assets operates 15 data centers across North Carolina, South Carolina, and Texas in the United States, plus facilities in Sweden — representing 1.3 gigawatts of available power capacity with energized and approved grid connections. Multiple Texas sites were already described by GDA as "hyperscaler-grade," a meaningful engineering distinction: facilities built to traditional data center specifications require significantly less conversion work than standard mining sheds when it comes to meeting AI workload requirements.

SWI Group founder and CEO Max-Hervé George framed the acquisition as the result of a long-held infrastructure thesis. "Power connectivity is the most valuable commodity in digital infrastructure today, and converting legacy cryptocurrency mining infrastructure to AI and high-performance computing is the best and highest use of these assets," George said. "We have been investing in power-connectivity since 2020. This is what that thesis looks like at scale."

With the GDA acquisition complete, AiOnX now controls a combined 3.6 gigawatts of data center capacity spanning the United States and Europe — a transatlantic footprint that positions it among the larger independent AI infrastructure platforms operating outside the hyperscalers themselves.

What Bitcoin Mining Infrastructure Actually Converts to AI — and What It Does Not

The phrase "conversion" understates the engineering work involved. A Bitcoin mining facility runs on ASIC chips — application-specific integrated circuits designed exclusively to solve the proof-of-work calculations that validate the Bitcoin blockchain. Those chips have no role in AI workloads. Every ASIC in every GDA facility will need to be removed and replaced entirely with GPU hardware: NVIDIA H100, H200, or GB200 processors, or AMD MI300X chips.

The hardware swap is only the beginning. AI training workloads distribute computation across many GPU nodes simultaneously, requiring high-speed interconnects — InfiniBand or 100 Gigabit Ethernet — that standard Bitcoin mining facilities do not carry. Storage infrastructure must be rebuilt: AI training requires high-throughput NVMe clusters and object storage pipelines, whereas mining requires almost none. Software stacks must also be rebuilt entirely around frameworks like PyTorch, TensorFlow, Kubernetes, and Slurm.

Power density requirements have escalated dramatically. Standard enterprise data centers historically ran at around 8 to 10 kilowatts per rack. Bitcoin mining facilities can handle 30 to 50 kilowatts per rack. NVIDIA's newest GPU clusters — the GB200 NVL72 — require more than 132 kilowatts per rack, which demands direct liquid cooling to the chip. Air-cooled mining halls cannot support that density without significant additions: liquid cooling systems, UPS units, battery backup, generator sets, and additional high-bandwidth networking.

Industry estimates for a full AI-ready conversion run between $8 million and $11 million per megawatt. At 1.3 gigawatts, that implies conversion capital expenditure in the range of $10 billion to $14 billion across GDA's full portfolio — a figure not included in the $500 million acquisition price, and one that will shape how quickly and in what sequence AiOnX brings the sites online. The hyperscaler-grade Texas facilities are the logical first movers.

AI Data Center Capacity Is Not Keeping Up With Demand

Gartner has projected that power shortages will restrict 40% of AI data centers by 2027. The cost of securing data center capacity across Europe's five largest markets — Frankfurt, London, Amsterdam, Dublin, and Paris — is forecast to rise by 12% in 2026, according to CBRE.

Mining sites offer something that purpose-built AI campuses do not: speed. Building a new data center from land acquisition to energization can take more than four years in the United States, in part because electrical grid connection queues have grown to multi-year lengths. Mining facilities already have energized substations, approved grid connections, and operational physical infrastructure. The conversion timeline for a hyperscaler-grade mining site is measured in quarters, not years.

GDA's Swedish facilities offer an additional strategic asset. Sweden's stable power infrastructure and renewable energy mix have made the Nordic region increasingly attractive for data center investment at a time when European power availability has become a competitive variable — particularly as costs in Germany, France, and the United Kingdom escalate.

AiOnX's Broader European Build-Out

The GDA deal is one piece of a rapidly expanding transatlantic platform. AiOnX is developing a data center campus at the Kildare Innovation Campus in Leixlip, County Kildare, Ireland, which has already been leased in full to a major US hyperscaler that Business Post and Irish EPA filings identify as Amazon Web Services. The entire 179-megawatt campus is under lease, with the first 16 megawatts of capacity scheduled to begin generating revenue in late 2026 under a 20-year agreement.

Beyond Dublin, AiOnX has projects underway in Cambridge, UK; Milan, Italy; Varde, Denmark; and Madrid, Spain, for a combined European pipeline of 2.3 gigawatts. The company is a wholly owned subsidiary of SWI Group, itself formed last year through the merger of Stoneweg and Icona Capital. SWI Group separately agreed in February 2026 to acquire a majority stake in Polarise, a German AI infrastructure platform and NVIDIA Cloud Partner that recently launched the first industrial-scale AI factory in Germany in partnership with Deutsche Telekom and NVIDIA.

How This Fits the Broader Industry Pattern

AiOnX is not the first to make this move at scale. CoreWeave, now one of the largest GPU cloud providers in the world, traces its origins to Bitcoin mining before repositioning as an AI infrastructure platform. IREN signed a five-year, $9.7 billion agreement with Microsoft in November 2025 to lease GPU-powered data center capacity built on NVIDIA GB300 processors at its Texas campus. Applied Digital, Cipher Mining, and Bitfarms have all announced pivots of varying scale.

What distinguishes the GDA deal is its scope and structure: 1.3 gigawatts of existing, energized capacity acquired in a single transaction, with a European operator bringing hyperscale leasing experience from its Dublin campus. The conventional pattern has been individual mining operators converting individual sites over time. This is a portfolio conversion executed as one move.

One piece of relevant background: in September 2025, the FTX Recovery Trust filed a $1.15 billion lawsuit against Genesis Digital Assets — the mining company being acquired here, which is distinct from the bankrupt lender Genesis Global Capital — alleging that between 2021 and 2022, Alameda Research invested more than $1 billion in GDA using misappropriated FTX customer funds. GDA has not admitted wrongdoing; the lawsuit is ongoing. SWI Group did not address the litigation in its announcement.

Frequently Asked Questions

Why are Bitcoin mining sites useful for AI data centers?

Bitcoin mining facilities were built for high-density power consumption, continuous thermal load, and 24/7 operation — the same underlying requirements that AI GPU clusters demand. They typically have energized substations with direct grid connections that can take more than four years to secure for a purpose-built facility. Those permits, power agreements, and physical footprints transfer directly to AI operators, bypassing the bottleneck that currently restricts new data center construction. Mining sites are not plug-and-play for AI, but they are a faster starting point than a greenfield site.

What engineering work does converting a crypto mining site to AI compute actually require?

The hardware replacement alone is substantial: ASIC chips used for Bitcoin mining must be completely removed and replaced with AI GPUs. Networking infrastructure must be overhauled to support high-speed interconnects needed for distributed AI training — InfiniBand or 100 Gigabit Ethernet. Cooling systems must be upgraded to handle the power densities of modern GPU clusters, which can exceed 132 kilowatts per rack and require liquid cooling directly to the chip. Power redundancy systems — UPS units, battery backup, generator sets — must also be added. Industry estimates for a full AI-ready conversion run between $8 million and $11 million per megawatt.

What is a neocloud, and how does AiOnX fit into that market?

A neocloud is an AI-first cloud infrastructure provider that specializes in GPU compute for machine learning training and inference — purpose-built for AI, unlike the general-purpose hyperscalers such as AWS, Azure, and Google Cloud. Neoclouds typically offer bare-metal GPU access with transparent per-hour pricing and no broad managed services catalog. With 3.6 gigawatts of capacity spanning the United States and Europe after this deal, AiOnX is positioned to operate at hyperscale-adjacent scale as a neocloud provider, leasing capacity primarily to the hyperscalers themselves.

How constrained is AI data center capacity right now?

Gartner has projected that power shortages will restrict 40% of AI data centers by 2027. Of roughly 140 large-scale US data center projects representing approximately 12 gigawatts of planned capacity scheduled to come online in 2026, only about a third are currently under construction. In Europe, securing data center capacity across the five major markets is projected to cost 12% more in 2026 than in 2025, according to CBRE. Grid connection queues in the US now extend more than four years, making sites with existing substations and approved connections — like former mining facilities — worth a significant premium above their construction value.