The company building Claude, one of the most closely watched AI labs on the planet, may be borrowing money at junk rates. That’s the takeaway from a report that Blackstone has changed its mind about Anthropic’s credit standing, and the change matters more than it looks.
According to TechCrunch, citing the Financial Times, Blackstone told investors in a recent $63 billion financing deal for Anthropic’s chips that it expects the lab to receive a high junk rating after its IPO. Blackstone anchored that deal. The firm now expects a double B rating, the top tier of speculative grade. Earlier, Blackstone had said the opposite: that Anthropic would reach investment grade, the low-risk category for lenders, before 2027.
Two caveats. This is a forecast, not a rating, and the agencies haven’t spoken. Anthropic and Blackstone both declined to comment. But Blackstone is not a bystander. It has money on the table, so a downgrade of its own expectations is a signal worth taking seriously.
Why does the rating matter? Because Anthropic’s business runs on spending. The company has signed large compute deals over the past 12 months, and investors have grown nervous about it. A double B rating isn’t catastrophic. Plenty of large companies borrow at that level. But it means higher interest costs and a narrower pool of buyers, since many institutional investors are restricted to investment-grade debt. For a company that needs fresh capital constantly, that adds up.
The timing is awkward. Anthropic filed to go public earlier this week, and the IPO has already set off speculation about what it means for the wider AI industry. A credit story that turns sour right as public-market investors begin reading the filing is not what any CFO wants.
Compare this with the rivals. OpenAI leans on Microsoft and a large pile of private capital. Google and Meta fund their AI buildouts from enormous advertising cash flows, and both carry strong credit. Anthropic has no such parent. It has revenue growth, big backers, and a lot of commitments. That makes it more exposed when debt markets get picky.
Is this noise? Mostly not. The rating doesn’t change how good Claude is, and developers won’t see it in their API bills tomorrow. But the cost of capital shapes how fast a lab can train models and buy chips. If financing gets pricier, expect pressure on pricing, tighter spending, or more equity raises. Watch the actual agency ratings once the IPO closes. That’s when the guess becomes a fact.
