Anthropic signs $19 billion AI infrastructure lease with TeraWulf

TeraWulf announced on July 6, 2026, at 8:00 a.m. EDT that Anthropic has signed a 20-year lease for a purpose-built AI infrastructure campus in Hawesville, Kentucky. The deal is expected to generate about $19 billion in contracted lease revenue over its initial term and gives Anthropic access to a large dedicated compute site as demand for Claude and other frontier AI systems keeps rising.

The campus, called Justified Data, is planned for approximately 401 MW of critical IT load. TeraWulf says capacity will come online in phases, with initial service expected in the second half of 2027 and a full ramp to 401 MW by early 2028. The company also announced that it will sell its 50.1 percent stake in the Abernathy Joint Venture to an investor group led by Fluidstack, freeing capital from a roughly $450 million investment so it can focus on wholly owned AI infrastructure projects.

This is important because it shows how the AI race is moving from model announcements to industrial commitments. Frontier labs no longer need only chips in the abstract. They need power, land, cooling, grid connections, financing and long-term operators that can turn those pieces into reliable capacity. A 20-year, multi-hundred-megawatt lease is a very different signal from a short cloud contract. It suggests Anthropic is planning for years of heavy training and inference demand, not just another product cycle.

For AI users, there is no immediate change inside Claude. The practical effect is longer term: more dedicated capacity could support higher usage limits, steadier availability and potentially more predictable pricing if the buildout lands on schedule. For makers and businesses, the message is that access to frontier AI may increasingly depend on infrastructure deals signed years before the model reaches a user interface.

For the market, the deal reinforces a broader shift. Former crypto and data-center companies are repositioning around AI workloads, while AI labs are locking in power like strategic supply. The bottleneck in AI is becoming physical as much as technical.