SWI Group pivots 80% of capital into transatlantic AI infrastructure
SWI Group, the Euronext Amsterdam-listed investment group co-founded by Max-Hervé George and Jaume Sabater, has confirmed a sweeping strategic reorientation: more than 80% of the group's capital is now deployed into a transatlantic digital-infrastructure platform with a combined power capacity exceeding 4 GW, and the group intends to push that allocation above 90% over time. The announcement signals one of the more decisive balance-sheet pivots by a listed European investment vehicle into AI compute infrastructure to date.
Central to the shift is the completion of SWI's acquisition of a controlling stake of over 70% in Genesis Digital Assets (GDA), advised exclusively by Morgan Stanley. GDA will be rebranded SWI Digital and will serve as the group's US-focused digital-infrastructure arm, bringing an energised, grid-connected land bank that gives SWI immediate access to what the group describes as the world's largest and fastest-growing market for AI and high-performance-computing capacity.
Building a vertically integrated compute stack
SWI's European flank is anchored by AiOnX, a portfolio of hyperscale, AI-ready data-centre campuses under development across Ireland, the United Kingdom, Denmark, Spain and Italy. One site has already been secured by a leading hyperscale tenant, though the group has not named the counterparty. Together, AiOnX and SWI Digital form a two-continent land-and-power base from which SWI intends to move up the value chain.
The group plans to develop a proprietary GPU-as-a-service platform in-house, combining the energised sites of both platforms with an HPC compute layer. The strategic logic is vertical integration: by owning land, grid-connected power, data-centre capacity and the compute layer above it, SWI aims to capture margin at each rung of the AI-infrastructure stack rather than acting purely as a landlord leasing capacity to hyperscalers. A previously announced partnership with Polarise has been restructured; instead of acquiring a majority stake in the firm, SWI will provide financing to accelerate Polarise's independent expansion.
The convergence capital angle
SWI's pivot illustrates a broader reallocation trend playing out across listed and private European investment vehicles. As hyperscaler capex commitments from Microsoft, Google and Amazon have validated the demand signal for AI infrastructure at scale, a new class of capital allocator has emerged: listed investment groups and family-office-style balance sheets that can move faster than infrastructure funds constrained by LP return cycles, deploying into pre-hyperscale and co-location capacity before institutional consensus forms.
The 4 GW platform SWI is assembling sits at the intersection of several macro forces. Power procurement is the binding constraint in European AI infrastructure, making grid-connected land banks in Ireland, Denmark and the Iberian Peninsula strategically scarce assets. In the US, SWI Digital's pre-energised land bank addresses the same bottleneck. The decision to layer GPU-as-a-service onto owned infrastructure also positions SWI to compete directly with cloud-native AI compute brokers, a market where pricing pressure is intensifying as supply from hyperscalers and independents alike comes online.
Beyond digital infrastructure, SWI retains diversified exposure through European industrial and logistics real estate via Singapore-listed SERT, US multifamily residential through Swiss-listed Varia US, and an emerging allocation to culture, sport and entertainment. George noted that the group's listing "gave us the balance-sheet firepower and the agility to back the trends we believe will define the next decade." The forward-looking framing is notable: SWI expects double-digit balance-sheet growth in 2026, a projection the group has not yet underpinned with granular revenue or EBITDA guidance.
For cross-sector investors, the SWI story is a proxy for the structural question now facing every capital allocator with exposure to AI: whether the durable returns in the AI infrastructure cycle accrue to the hyperscalers and chip designers at the top of the stack, or to the owners of the physical layer underneath them.