SWI Group reports strong H1 2026 results as it accelerates transformation into a global AI infrastructure and compute platform

SWI Group reported €631.6 million in H1 2026 profit and €2.3 billion in adjusted NAV, up 53% from December 31, 2025, primarily reflecting the valuation of its Genesis Digital Assets investment. Following period-end acquisition steps, SWI obtained control of rebranded SWI Digital with roughly 1.2 GW of secured US grid connections, complementing AiOnX's approximately 2.3 GW European planned capacity. The company is pivoting toward an integrated AI infrastructure platform, targeting digital infrastructure at more than 90% of assets by 2027 and discussing hyperscaler/AI-developer offtake agreements with aggregate potential value in the tens of billions of dollars.

Analysis

The investable question is not the stated pipeline but whether SWICH can turn grid interconnection rights into creditworthy, financed take-or-pay contracts before the AI-capex cycle normalizes. The reported earnings and NAV expansion appear substantially mark-driven rather than cash-flow evidence, leaving the equity exposed to a discount if project-level funding costs, conversion capex, or customer prepayment terms prove less favorable than assumed. A US listing could improve access to AI-infrastructure capital but is also a likely dilution overhang in the next 3-9 months.

Vertical integration is strategically attractive only if SWI can procure GPUs and power at a cost below what hyperscalers can internally achieve. NVIDIA is the clearest indirect beneficiary if deployments materialize, but preferred-partner status is not equivalent to committed hardware allocation or end-customer demand; the market should not capitalize it as an offtake agreement. The more probable near-term value capture lies with liquid power-and-data-center comparables such as EQIX, DLR and VRT rather than a small, asset-mark-dependent developer.

A non-obvious risk is that converting mining-oriented sites to AI capacity requires far more than available power: fiber redundancy, liquid cooling, building shells, transmission upgrades and high-utilization GPU customers. That can shift economics from low-capex powered-land valuation toward a capital-intensive cloud model with rapid hardware depreciation. Conversely, a signed hyperscaler lease with customer-funded buildout would materially de-risk financing and justify a valuation rerating within 1-3 months; absent that disclosure, the current narrative is not sufficient for a high-conviction position.

Consensus AI enthusiasm may underprice the bottleneck value of secured power, but it also overstates the scarcity premium for capacity that is not operationally deliverable. The thesis is falsified by a delayed contracted-capacity announcement, project-finance spreads widening materially, a capital raise below NAV, or evidence that GPU utilization cannot support returns above the cost of capital over the next 6-18 months.

Key Decisions for Investors

• Do not initiate a directional SWICH position on the release alone. Place on event-driven watch for a disclosed binding offtake agreement specifying MW, term, customer credit quality, capex responsibility and financing; absent these details, treat the reported opportunity value as non-investable pipeline.
• If SWICH trades at a substantial premium to independently verified NAV before contracted capacity or financing is disclosed, consider a small tactical short over a 1-3 month horizon. Cover on a binding hyperscaler contract with customer prepayments or a US listing priced above book value; risk is a momentum-driven AI-infrastructure rerating in a thinly traded security.
• Maintain NVDA exposure rather than adding it specifically for this development. Reassess only if SWI discloses a material committed GPU order or capacity deployment schedule; partner-network designation alone has immaterial impact on NVDA revenue versus its existing hyperscaler order book.
• For broad AI-infrastructure exposure, prefer a liquid long basket of VRT and DLR versus a short basket of speculative powered-land developers where feasible. The pair targets the spread between proven backlog/operating cash flow and uncontracted interconnection-right valuations over 6-12 months; reduce if long-duration rates decline sharply or multiple new hyperscaler leases validate developer pipelines.