Anthropic's $1.3B Mezzanine Loan for Hubbard Texas Data Center: What AI Builders Should Know

Eagle Point Credit Management is providing a roughly $1.3 billion mezzanine loan to Nexus Data Centers for a 2,900-acre AI data center campus in Hubbard, Texas, with Anthropic as the anchor tenant. The loan closes the capital structure for the first phase of a project that carries a total financing package of about $16 billion, led by Morgan Stanley with Google guaranteeing the senior debt. For AI builders, the deal illustrates how large labs secure compute capacity through project finance rather than owning facilities, but it also raises questions about leverage and local resource strain.

Eagle Point's $1.3B Mezzanine Loan Completes First Phase of Hubbard Campus

The mezzanine loan sits atop the capital stack, carrying higher risk and higher potential returns than the senior debt. Eagle Point is the single largest investor in this tranche, which it structured as a "HoldCo" loan. The financing allowed Nexus to reach a final investment decision in under a year, a timeline the firm called "a truly remarkable achievement."

The Hubbard campus will include a gas-fired power plant and operate as a behind-the-meter site, meaning it generates its own electricity rather than drawing from the grid. The senior portion of the package, roughly $15 billion, is led by Morgan Stanley and guaranteed by Google. Anthropic won a competitive process to become the primary tenant.

How the Deal Structure Affects AI Builders

For AI labs like Anthropic, this financing pattern provides access to a large new compute site without owning or paying for the buildings up front. Anthropic has used similar structures before, including a partnership with Macquarie. The model lets AI companies lock in capacity while shifting capital intensity to lenders and developers.

Google's guarantee on the senior debt is a key risk mitigator. It makes lenders more comfortable funding large sums, but it also concentrates financial exposure with fewer parties. If the tenant or backer runs into trouble, the risk sits with Google and the guaranteed entities. This structure is becoming common across the industry, with other labs signing multibillion-dollar deals for compute capacity.

Practical Implications: Compute Access, Refinancing, and Energy Strategy

The speed of this deal matters. Nexus reached its final investment decision in under a year, which is fast for a project of this scale. For AI builders, that means capacity can come online relatively quickly when the financing is structured correctly.

However, the project is not fully funded yet. Nexus may sell high-yield bonds later this year to refinance the $15 billion senior portion, a step that would require a credit rating. If credit markets tighten, that refinancing could become more expensive or delayed, affecting the overall cost of capital.

The behind-the-meter power arrangement reduces strain on the Texas grid but concentrates energy risk with the project and its backers. The campus will rely on its own gas-fired plant, which insulates it from grid volatility but exposes it to fuel costs and regulatory changes. Texas officials have raised questions about the power and water demands of these large data centers, which could affect permitting and public perception.

Caveats and Open Questions

Several details remain uncertain. The financing terms and timelines could change as the project progresses. The refinancing plan depends on credit market conditions, which are unpredictable. Local scrutiny over energy and water use in Texas may slow development or add conditions. The deal also adds to a growing pile of data-center debt, raising broader questions about how much of the AI boom is running on borrowed money. Whether the Hubbard bet pays off will depend on demand holding up long enough to repay the borrowing.