Anthropic’s $800B AI Gamble Gets a Junk-Grade Curveball Before Its IPO

Blackstone has reversed its earlier forecast and now expects Anthropic to receive a below-investment-grade credit rating following its planned IPO, citing feedback from major rating agencies. The warning raises concerns about the AI company’s borrowing costs as it prepares to spend more than $800 billion on computing infrastructure through 2045.

Anthropic, the company behind Claude, had previously been expected by Blackstone to obtain an investment-grade rating by year-end. Kroll Bond Rating Agency (KBRA) assigned Anthropic a BB+ rating in June, one notch below investment grade. Fitch, Moody’s, and S&P have not issued official ratings, although reports said in September that the agencies continued to view Anthropic as speculative grade despite backing from Goldman Sachs and Morgan Stanley. A final determination is expected after Anthropic’s Nasdaq debut, anticipated by late November.

The company completed a $63 billion financing arrangement called Project North Star, covering leases for 1.5 gigawatts of computing capacity using Google-designed chips. Broadcom guarantees more than $42 billion in senior loans, while $18 billion of junior debt is backed by Anthropic itself. Blackstone funds took about $9 billion of that junior portion at yields above 10%.

Blackstone’s revised outlook could limit Anthropic’s access to institutional lenders because some pension funds and insurers restrict investments in speculative-grade securities. Anthropic holds more than $112 billion in cash, and investors expect its IPO to raise roughly $100 billion more. Its prospectus describes the compute plan as “bold but sane,” according to the Financial Times.

SpaceX received investment-grade ratings from all three major agencies shortly after its June IPO. For Anthropic, the prospective rating could determine how readily it accesses debt markets and at what cost as it finances its infrastructure expansion.