Google to Back $15 Billion Loan for Anthropic's Texas Data Center, Report Says

A data center developer working for AI startup Anthropic is in the final stages of securing $15 billion in financing for a new data center campus and power plant in Texas, backed by significant financial guarantees from Google, The Wall Street Journal reported on Thursday.

Details of the Financing

According to the report, a group of banks led by Morgan Stanley is set to provide the loan to the developer, Nexus Data Centers. The financing package reportedly includes a $14 billion bridge loan and a revolving credit facility to fund the construction of the facility in Hubbard, Texas.

A key component of the project is a dedicated natural gas power plant with 1.6 gigawatts of generating capacity. This underscores the immense and growing energy requirements for training and operating large-scale artificial intelligence models.

Google's Strategic Role

Google's involvement is critical to the deal's structure. The technology giant has reportedly issued guarantees covering billions of dollars of Anthropic's lease and power payment obligations, mitigating risk for the lenders should the AI startup fail to meet its commitments.

Ad

In return for its financial backstop, Google will receive an equity stake of approximately 20% in the data center and power project. The report also noted that Anthropic will deploy Google's custom Tensor Processing Units (TPUs) at the site, with a separate vendor financing agreement with Broadcom covering the cost of the chips.

Market Implications

This arrangement highlights an emerging financing model where major technology firms use their balance sheets to underwrite the massive capital expenditures required for AI infrastructure. It allows AI developers like Anthropic to secure the vast computing resources they need to compete.

The deal is part of Anthropic's broader strategy to lease data centers directly, aiming to secure at least 10 gigawatts of capacity in the coming years. This model provides more direct control over computing resources compared to renting server capacity from public cloud providers.