Anthropic, the company behind the Claude models, has entered into a 20-year, $9.1 billion agreement with Riot Platforms for 191 megawatts of capacity at Riot’s Rockdale campus in Texas. The deal underscores how the rapid expansion of artificial intelligence is reshaping demand for power and computing infrastructure, and creating a new use for facilities built during the cryptocurrency mining boom.
As Bloomberg reported, full delivery is expected by mid-2028, with extension options that could raise the total value to approximately $16.1 billion. The deal is less a one-off transaction than a clear illustration of how AI’s escalating compute needs are increasingly overlapping with the infrastructure originally built for digital asset mining.
The Surging Demand for AI Compute
Modern AI models require extraordinary volumes of computation for both training and inference. Each new generation of large language models multiplies the number of parameters and the volume of data processed, translating directly into higher power consumption and denser clusters of specialized chips.
Leading AI laboratories, including Anthropic and OpenAI, have responded by locking in capacity years in advance, often through multi-year or multi-decade contracts that guarantee access to electricity and cooling at scale.
This pressure is industry-wide. Traditional cloud providers cannot expand fast enough to meet the combined needs of model developers and enterprise customers. As a result, AI companies are turning to alternative sources of power and real estate—sites that already possess robust electrical interconnects, industrial zoning, and proven operational experience with high-load environments.
Riot Platforms’ 191-megawatt commitment, equivalent to the electricity demand of tens of thousands of homes, represents only one piece of a much larger infrastructure race currently underway.
Riot Platforms Posts Higher Revenue Amid Data Center Pivot in Q2
Riot Platforms reported second-quarter 2026 revenue of $174.2 million, a 14% increase from $153.0 million in the same period a year earlier, driven by growing contributions from its data center and engineering segments even as bitcoin mining revenue declined. Data center revenue reached $23.2 million, reflecting the completed delivery of the initial 25 megawatts to AMD, while engineering revenue rose sharply to $37.3 million.
The company produced 1,587 bitcoin during the quarter and ended the period with more than $1.2 billion in liquid assets, including 11,380 bitcoin and $548.9 million in cash.
Despite the top-line growth, Riot recorded a net loss of $237.2 million, or $0.68 per diluted share, compared with net income of $219.5 million a year earlier. The results were released alongside the announcement of the landmark 191-megawatt data center lease, underscoring management’s emphasis on shifting toward long-term AI and high-performance computing contracts as a more stable complement to its traditional mining operations.
Compared to Q1 results, Riot Platforms’ second-quarter 2026 revenue marked a sequential growth of roughly 4% from $167.2 million in the first quarter as data center contributions continued to expand.
Parallel Energy Dynamics with Bitcoin Mining
Bitcoin mining and AI training share fundamental economic and technical characteristics. Both are energy-intensive forms of continuous computation. Miners expend electricity to perform cryptographic hashing; AI systems expend electricity to perform vast numbers of matrix operations. In each case, the dominant cost is power, and the dominant operational challenge is delivering that power reliably and at high density while managing heat.
Bitcoin mining firms were early to identify regions with surplus or low-cost electricity and to develop the facilities, cooling systems, and grid relationships needed to operate at scale. Many of those same assets—large campuses, high-capacity substations, and experience running dense compute loads—are now proving highly suitable for AI workloads.
Converting or expanding mining infrastructure for GPU clusters can be faster and more capital-efficient than building new data centers from scratch. The Rockdale site exemplifies this advantage: it already possessed the electrical backbone that would take conventional developers years to assemble.
Read: Bitcoin Mining Squeeze: Nearly 23% of Mainstream Miners Are Operating at Loss
Convergence of Two Energy-Hungry Sectors
The growing collaboration between AI developers and Bitcoin miners signals a structural shift. Mining companies are diversifying into stable, long-duration compute leases, while AI firms gain access to power and facilities that might otherwise remain scarce. Long-term contracts of this magnitude provide miners with predictable revenue streams independent of cryptocurrency price cycles and give AI laboratories greater certainty of supply in a constrained market.
Broader implications are already visible. Competition for electricity in key regions is intensifying. Hybrid facilities capable of supporting both mining and AI workloads may become more common. Investors and grid planners must now treat the two industries as closely linked consumers of power rather than entirely separate sectors.
As models continue to scale and inference demand rises globally, the ability to secure and deliver large blocks of electricity will remain as critical as algorithmic progress itself. The overlap between AI compute and Bitcoin mining infrastructure is no longer theoretical—it is becoming a defining feature of the next phase of technological expansion.
