Hut 8's Texas power site sits inside Anthropic’s $35 billion AI deal

Hut 8’s Texas power site has found itself at the epicenter of artificial intelligence’s relentless search for high-density compute power, anchored directly within Anthropic’s massive $35 billion infrastructure expansion plan. The digital asset infrastructure provider has locked in two long-term leases valued at a staggering $19.6 billion—a figure that represents more than 260 times the company’s most recent quarterly revenue. This sheer scale underscores a fundamental shift in the economics of digital infrastructure, as hyperscalers and frontier AI labs aggressively outbid traditional enterprise tenants to guarantee access to energized data-center capacity. For institutional investors and site developers, the development marks a definitive inflection point in the convergence of crypto mining assets and high-performance computing (HPC). Tech giants are encountering severe bottlenecks in their push for gigawatt-scale AI training clusters, driven primarily by multi-year waiting lists for power grid interconnections and electrical transformers. Bitcoin miners like Hut 8, which spent years securing massive power purchasing agreements across grid operator ERCOT’s Texas territory, now possess the one asset AI firms cannot immediately build with capital alone: speed-to-market backed by ready, high-voltage power access. The economics of these long-term commitments fundamentally rewrite the valuation framework for power-focused operators. Moving away from the inherent volatility of Bitcoin block rewards and transaction fees, operators transitioning capacity toward AI workloads are securing fixed, annuity-style cash flows backed by investment-grade balance sheets. A $19.6 billion lease profile provides Hut 8 with unprecedented financial visibility, drastically lowering its cost of capital and setting a benchmark for how grid-adjacent real estate will be priced moving forward. Developers across the Sun Belt are taking note, rapidly recalibrating their pipeline strategies away from standard cloud co-location toward high-density, liquid-cooled architectures customized for trillion-parameter model training. Yet this dynamic extends beyond a single corporate windfall; it signals a structural reallocation of energy resources across North America. As Anthropic and competing frontier AI labs scale their compute requirements toward megawatt and gigawatt thresholds, power capacity has superseded chip availability as the primary constraining factor in modern technology development. Capital markets are already adjusting, assigning premium valuations to operators capable of converting energized hash-rate footprint into AI-ready infrastructure. For developers and institutional funds watching from the sidelines, the message is unmistakable: control of raw power capacity is now the ultimate moat in the global AI race.