In a historic move to secure long-term computing power for its Claude AI models, AI research firm Anthropic has signed a $9.1 billion, 20-year infrastructure agreement with Riot Platforms. Disclosed on August 12, 2026, the deal highlights how frontier AI developers are moving aggressively to guarantee physical datacenter capacity and energy access over multi-decade horizons.
Securing 191 Megawatts of AI Datacenter Capacity
Under the terms of the 20-year agreement, Anthropic will secure 191 megawatts (MW) of dedicated high-density computing capacity from Riot Platforms’ flagship datacenter facility in Texas.
The facility will house specialized high-performance GPU clusters engineered specifically for training next-generation Claude models and serving large-scale enterprise inference workloads. The long-term structure locks in predictable energy costs and hosting real estate at a time when global cloud providers face severe power grid constraints.
Key Takeaways for the Cloud & AI Ecosystem
Industry analysts emphasize that multi-decade compute contracts signal a fundamental shift in how frontier AI labs manage supply chain risk:<
- Transition from Cloud Bursting to Physical Real Estate: Frontier AI labs are increasingly bypassing traditional short-term cloud leasing in favor of direct, multi-decade power and facility capacity lock-ins.
- Energy as the Primary AI Bottleneck: Securing 191 MW of power underscores that access to stable, high-capacity electrical grids is now the dominant limiting factor in AI scaling.
- Long-Term Capital Commitment: The $9.1 billion commitment reflects deep institutional confidence in sustained enterprise demand for generative AI and agentic workflows through the 2030s and 2040s.