Google is backing a major expansion of Anthropic’s computing infrastructure in Texas, connecting the AI developer with a planned data center buildout involving infrastructure operator TeraWulf. The project has been valued at about $15 billion, while the underlying long-term lease could generate roughly $19 billion in contracted revenue for TeraWulf.
TeraWulf disclosed a 20-year agreement under which Anthropic would lease approximately 401 megawatts of capacity at the company’s Justified Data campus. The substantial power allocation reflects the scale of infrastructure now required to train and operate advanced AI models, which rely on large clusters of specialized processors as well as extensive power, cooling and networking systems.
Google’s support includes a $1.8 billion lease backstop and an equity stake in TeraWulf. A lease backstop can reduce a data center developer’s financial exposure by providing additional assurance around a tenant’s payment obligations. That support may make it easier for TeraWulf to finance construction, although the project still carries the execution risks associated with developing power-intensive computing facilities.
The arrangement deepens the infrastructure ties between Google and Anthropic. Google is already a major investor and cloud partner to the AI company, even as Anthropic also uses infrastructure supplied by other providers. Supporting additional capacity outside Google’s own data center fleet gives Anthropic another route to secure the computing resources needed for its Claude model family and related services.
For Google, the transaction extends a strategy that combines direct investment in AI developers with cloud services, custom chips and financial commitments supporting new infrastructure. Securing access to computing capacity has become a central competitive concern for large AI companies because shortages in suitable sites, electrical connections and advanced accelerators can constrain how quickly they deploy new systems.
TeraWulf’s involvement also illustrates the transition taking place among some digital infrastructure businesses that previously concentrated on cryptocurrency mining. Both cryptocurrency operations and AI data centers require access to large amounts of electricity, but converting or expanding sites for AI workloads demands different hardware, networking, cooling and reliability standards. AI tenants generally expect data center systems designed for dense accelerator clusters and continuous enterprise workloads.
The headline figures describe different parts of the arrangement rather than a single payment. The approximately $15 billion figure refers to the reported value of the Texas infrastructure project, while the roughly $19 billion total represents contracted lease revenue expected across the 20-year term. Those revenues would be collected over time and remain dependent on the project being completed and the lease operating as planned.
Building 401 megawatts of capacity will require more than installing servers. Large AI campuses need high-voltage electrical infrastructure, redundant systems, cooling equipment and high-speed links capable of connecting thousands of accelerators. Construction schedules can also depend on utility interconnections, equipment availability, financing and local permitting.
The deal gives Anthropic a path to a large block of dedicated capacity while offering TeraWulf a long-duration tenant commitment. Google’s backstop provides another layer of financial support, but the next phase will depend on TeraWulf delivering the campus infrastructure and Anthropic bringing its computing systems online over the planned development schedule.