Infrastructure financier IREN's ambitious $25 billion to $30 billion capital allocation toward artificial intelligence compute capacity appears to have triggered investor caution, according to recent analysis from Bernstein. The scale of the deployment represents one of the largest bets on AI infrastructure buildout by a single entity, but the sheer magnitude may have signaled execution risk or oversaturation concerns to market participants watching the sector consolidate around a handful of mega-players.
The numbers underlying IREN's strategy reveal a sophisticated play on the compute shortage narrative that has dominated venture and infrastructure discourse since 2023. The firm has already locked in $4 billion in annualized recurring revenue across its contracted cloud capacity for 2026, demonstrating genuine demand signals from hyperscalers and enterprise customers. More tellingly, recent contract negotiations have yielded over $20 million in annual revenue per IT megawatt—a metric that reflects pricing power and the premium buyers currently place on available GPU and processing capacity in a supply-constrained market.
This revenue intensity matters because it reveals whether IREN can justify its infrastructure spend through productive utilization. The per-megawatt figure suggests customers view compute access as mission-critical and are willing to pay accordingly, which validates the underlying thesis. However, investor skepticism may stem from broader concerns about competitive dynamics. With major cloud providers—Amazon, Google, Microsoft—racing to build proprietary AI infrastructure alongside purpose-built startups and traditional data center operators, questions arise about whether independent infrastructure players can maintain pricing power or secure long-term contracts as capacity eventually normalizes.
The market's apparent wariness also reflects uncertainty around deployment timelines and utilization ramps. Building out $25 billion to $30 billion in infrastructure requires navigating supply chain constraints, regulatory approval processes, and the technical complexity of optimizing power delivery and cooling for high-density AI workloads. If IREN faces delays or encounters lower-than-projected utilization rates in early phases, the investment thesis deteriorates quickly. The 2026 contracted ARR provides a helpful floor, but investors remain acutely aware that hyperscaler spending cycles can shift unexpectedly—a risk that becomes acute when capital commitments of this magnitude are on the line. Whether IREN can execute this buildout while maintaining unit economics will likely define infrastructure investing throughout the decade.