Anthropic, TeraWulf Sign $19 Billion Computing-Power Deal

Training and running generative AI models requires vast data-center capacity, advanced chips and power, making access to compute a central battleground for leading developers. Anthropic has been expanding its infrastructure footprint as it competes with OpenAI, while UK startup Nscale is building large-scale AI data centers. The agreement also underscores the strategic importance of Nvidia’s Vera Rubin platform in the next wave of model development.

Anthropic agreed to spend about $45 billion with Nscale to secure computing resources powered by Nvidia’s latest Vera Rubin chips. Support under the agreement is expected to begin by the end of 2027, giving Anthropic additional capacity for training and serving its AI systems. The deal is the latest step in the company’s compute-buying push and marks a significant escalation in its cloud and data-center spending.

TeraWulf began as a bitcoin miner but has been redeploying its power access, land and data-center expertise toward high-performance computing infrastructure as demand for AI capacity accelerates. Anthropic, the developer of the Claude models, is seeking large blocks of long-duration compute. The lease gives TeraWulf a more predictable revenue stream, reducing its exposure to cryptocurrency cycles and cementing HPC leasing as a central business line.

TeraWulf said on July 6, 2026, that its Raylan Data unit signed a 20-year lease with Anthropic at the Justified Data campus in Hawesville, Kentucky. The initial term is expected to generate about $19 billion in contracted lease revenue and cover roughly 401 megawatts of critical IT load. Capacity is due to enter service in phases in the second half of 2027 and reach the full 401 MW by early 2028; Anthropic holds two five-year renewal options.