Roughly the continuous power demand of three large nuclear reactors, the transatlantic footprint positions the Amsterdam listed mid cap to bid for hyperscale and sovereign AI tenants.
SWI Group has signed a binding agreement to take majority control of Genesis Digital Assets, a US digital infrastructure operator, lifting the Amsterdam-listed company's combined AI and high-performance computing data-center capacity past 3.6 gigawatts, roughly the continuous power demand of three large nuclear reactors, or enough to run about 3.6 million homes.
The deal, announced 15 June 2026 via PRNewswire, moves SWI from minority shareholder to majority owner of GDA, a company SWI describes as one of the largest private digital infrastructure operators in the United States. GDA runs 15 sites, including several hyperscale campuses in Texas, with land carrying 1.3 GW of commissioned or approved grid connections. A gigawatt, for context, is the kind of continuous load a large city draws.
SWI already owns a European data-center platform called AiOnX, with a 2.3 GW pipeline across five hyperscale projects. Combined with GDA, the group is positioning itself as a vertically integrated, transatlantic owner-operator. In plain terms, that means SWI owns and runs the physical layer (land, power connections, buildings, cooling) rather than renting capacity from someone else. The stated strategy is to redirect GDA's assets toward high-performance computing and AI workloads, the parallel-compute layer used to train large AI models and run scientific simulations.
A third piece is in motion. SWI has a pending majority stake in Polarise, a group with NVIDIA cloud-service-provider status (a partner tier authorized to rent out NVIDIA GPU capacity) and a Deutsche Telekom/NVIDIA AI factory in Germany. If the Polarise deal closes, the three platforms would give SWI direct ownership of the physical infrastructure behind AI training runs in three jurisdictions: Texas, five European countries, and Germany.
The relevant question is what kind of operator can convert that footprint into hyperscaler contracts. Hyperscalers, the multi-hundred-megawatt, single-tenant or single-customer campuses run by Microsoft, Google, Amazon, and Meta, have historically preferred to build and operate their own data centers, financing the capital expenditure from their own balance sheets. Mid-cap aggregators like SWI are betting that not every AI workload fits that template, and that some hyperscalers, sovereign AI programs, and enterprise AI buyers will pay a premium for ready-made, grid-connected campuses in the right jurisdictions.
The 15 June announcement is silent on price, closing date, and the exact post-deal stake. The 1.3 GW and 2.3 GW figures are company-self-reported, not independently verified, and the "hyperscale" label is SWI's framing rather than a third-party operator's confirmation. SWI is also not disclosing customer concentration, leverage, or the contractual terms of the binding agreement. Until the Euronext Amsterdam filing or an independent confirmation lands, the 3.6 GW number should be read as a corporate claim, not an industry benchmark.
What to watch next: the Euronext Amsterdam regulatory filing for the GDA transaction, the closing date, the named hyperscale tenant on AiOnX (SWI has named one in prior disclosures), and the closing of the Polarise deal. If SWI files with concrete pricing, that single datapoint will let analysts convert a footprint claim into a market-position claim.