Nexus Data Centers is closing in on $15 billion in financing to build an AI campus in Hubbard, Texas, which Anthropic will lease, with Morgan Stanley leading the bank group behind the deal, according to CNBC.
The loan is going to Nexus, a data center developer few investors had heard of a year ago, and the reason banks are comfortable lending that much money traces back to a company that isn’t building anything on the site at all: Google.
Related: Anthropic clarifies stance on open-weight AI models
The debt sits with a developer, not with the AI lab
The financing package includes a $14 billion bridge loan and a revolving credit facility, structured for Nexus Data Centers rather than for Anthropic directly, according to Bloomberg.
That distinction matters more than the headline number. Anthropic’s own balance sheet, whatever its strength, isn’t what lenders are underwriting here.
Instead, Google agreed to backstop Anthropic’s obligations using its investment-grade credit rating, CNBC reported.
The guarantees cover four Anthropic data center leases and the power purchase agreements tied to an on-site natural gas plant with 1.6 gigawatts of capacity.
Google’s exposure was limited to the minimum amount lenders required to close the deal, the same report said.
A guarantee is cheaper than a check
Google didn’t need to write Nexus a check to make this project happen. It needed to lend its credit rating, and in exchange it is expected to receive roughly 20% equity ownership in the data center and power project, according to CNBC.
That’s a capital-efficient way to buy into AI infrastructure without tying up $15 billion in cash.
It also isn’t Google’s first exposure to Anthropic’s balance sheet. In April, Google agreed to invest up to $40 billion in Anthropic, starting with $10 billion in cash at a $350 billion valuation, according to CNBC.
Anthropic’s valuation has since climbed to $965 billion in a Series H round in May, Anthropic said.
Its run-rate revenue crossed $30 billion this year, up from roughly $9 billion at the end of 2025, Anthropic disclosed when it expanded its Google and Broadcom compute partnership.
Broadcom is involved on the hardware side too. Anthropic plans to deploy tensor processing units co-designed by Google and Broadcom at the Hubbard site, with a separate vendor financing agreement covering the chip costs, CNBC reported.
Google now plays four roles in this relationship
Google competes with Anthropic through Gemini. It also supplies Anthropic’s cloud and chip capacity, holds an equity stake as an investor, and now guarantees its landlord’s debt.
Few companies carry that much simultaneous exposure to a single rival.
Alphabet (GOOGL) shares slipped about 0.7% Thursday afternoon as the report circulated, reflecting immediate investor anxiety over Big Tech taking on contingent liability risk for a private AI lab’s real estate debt.
The Nvidia-OpenAI playbook looks familiar
This structure isn’t unique to Google and Anthropic. Nvidia is in talks to provide up to $250 billion in financing guarantees for a 10-gigawatt OpenAI data center campus in Ohio, CNBC confirmed in July.
Both arrangements exist for the same reason. Neither Anthropic nor OpenAI can borrow tens of billions of dollars on its own credit, so the chip or cloud partner steps in as the guarantor lenders actually trust.
The scale is different. Nvidia’s proposed backstop is more than fifteen times the size of Google’s.
But the mechanism is identical, and it’s becoming the standard way AI labs finance physical infrastructure without owning it outright.
More AI:
- Workers just sent AI companies an ultimatum
- Palantir CEO has a blunt verdict on OpenAI and Anthropic
- Elon Musk pulls no punches with AI rivals as Grok 4.5 debuts
Big Tech’s balance sheets are becoming AI’s credit market
The Hubbard campus lands at an awkward moment for that model. Five-year credit default swaps on Alphabet’s own debt have widened this year as bond investors price in more risk from the AI buildout, Bloomberg reported in its Credit Weekly note.
Guaranteeing a third party’s leases adds another contingent liability onto a balance sheet the market is already watching more closely.
That works as long as demand for AI compute keeps outpacing supply. If it doesn’t, the guarantees that make deals like this bankable today become the channel that transmits AI’s financial risk directly onto the world’s largest companies.
Investors tracking Alphabet’s earnings should start reading the footnotes on contingent obligations as closely as they read the revenue line.
Related: Anthropic spills the beans on reality of AI, jobs, and the economy
