SWI Group informa sólidos resultados en el primer semestre de 2026

SWI Group reported H1 2026 profit of €631.6 million and adjusted NAV of €2.3 billion, up 53% from December 2025, following its investment in Genesis Digital Assets, now renamed SWI Digital. After period-end, SWI increased its holding to roughly 70% of voting rights, securing control of a U.S. digital-infrastructure platform with about 1.2GW of grid connections; together with AiOnX, the group has approximately 3.5GW of European and U.S. capacity. SWI expects digital infrastructure to exceed 90% of total assets by 2027 and is pursuing long-term cloud and AI customer contracts with potential aggregate value in the tens of billions of U.S. dollars.

Analysis

The investable issue is execution financing rather than reported asset value. SWICH is attempting to re-rate from a capital allocator into an AI-infrastructure operator, but the economics depend on converting grid access into contracted, creditworthy load before committing GPU and fit-out capex. Until signed take-or-pay agreements, the asset uplift is substantially less reliable than recurring EBITDA; a US capital-markets transaction could also introduce dilution precisely when the equity is being valued on optionality.

NVDA is the clearest liquid read-through, but incremental demand from a single small-cap platform is immaterial to its estimates. The more relevant second-order implication is that redeploying former bitcoin-mining sites raises competition for scarce power interconnects, transformers, cooling equipment and electrical contractors; this is directionally supportive for Vertiv (VRT), Eaton (ETN), GE Vernova (GEV) and digital-infrastructure landlords with already contracted power. It is negative at the margin for marginal miners such as RIOT and CLSK if they must compete for power or accept lower-return AI conversion economics.

Over the next days, SWICH may trade on the NAV headline and NVIDIA-partner association, but neither establishes customer demand or funding terms. Over 1-3 months, signed hyperscaler contracts, disclosed MW, lease duration, customer credit quality and project-level returns are the needed catalysts. Over 6-18 months, conversion can create a higher-multiple recurring-infrastructure business only if utilization and power pass-through protect returns; construction delays, GPU procurement constraints, elevated rates, or a softer AI training-demand cycle would expose a levered development model.

Contrarian view: the market may be over-crediting a familiar "power-to-AI" narrative. Grid connections are valuable, but their monetization requires transmission reliability, permits, capital and customers, and not every mining facility meets hyperscale latency, redundancy and cooling standards. Treat the company’s prospective contract value as pipeline rather than backlog until counterparties and binding terms are disclosed.

Key Decisions for Investors

• No standalone SWICH long until the company discloses at least one binding contract with MW committed, term, pricing structure and committed financing. Reassess after the next disclosure; a capital raise at a material discount to reported NAV or failure to contract capacity within 90 days falsifies the near-term re-rating thesis.
• Express the infrastructure bottleneck through a 3-6 month basket long VRT and ETN, sized modestly, rather than NVDA: both capture power-density and cooling spend regardless of which cloud operator wins the workload. Exit or reduce if hyperscaler capex guidance weakens or order growth decelerates materially at either company.
• For a higher-beta relative-value expression, long VRT / short RIOT over 3-6 months. AI-site conversion increases demand for VRT’s equipment while raising the opportunity cost of power for miners; cover the short if bitcoin strength drives mining economics sufficiently high to preserve miner valuations despite power competition.
• Do not chase NVDA on this release. Maintain exposure only as part of the broader AI-capex book; the relevant confirmation is aggregate cloud-provider capex and GPU deployment guidance, not partner-network announcements from prospective operators.