Anthropic, the maker of the Claude chatbot, carried a BB+ issuer rating from Kroll Bond Rating Agency as early as June, according to documents reviewed by the Financial Times and disclosed on Friday, a previously unreported credit assessment that places one of the AI industry's most prominent builders a single notch below investment grade.
The rating sat unreported for four months while Anthropic's partners arranged tens of billions of dollars in chip financing around the company. Its disclosure now gives lenders and investors a formal credit benchmark for a firm whose infrastructure ambitions increasingly depend on borrowed money. For more context on this story, see our ongoing latest AI developments.
What the Rating Says
BB+ sits in speculative territory, the highest rung of what markets colloquially call junk status, one step below investment grade. KBRA assigned the BB+ issuer rating to Anthropic in June and also rated the junior tranche of a related financing at BB+ with a stable outlook, the FT reported.
The agency is not signaling distress. The newly disclosed assessment is not a downgrade of any prior rating, and KBRA's stable outlook suggests the agency expects Anthropic's credit profile to hold. What it establishes instead is a baseline: for the first time, there is a documented, agency-grade answer to the question of how a rating agency views the credit risk of a frontier AI lab with enormous revenue growth and enormous compute bills.
The June Financing It Shadowed
The rating was issued the same month that Apollo Global Management, alongside Blackstone and banking partners, announced a $35 billion commitment on June 9 to finance Broadcom's AI computing platform, an arrangement supporting more than one gigawatt of computing capacity intended for Anthropic.
That deal split lenders into two camps. Most of the money benefited from a guarantee from Broadcom, the chip designer at the center of the arrangement, while $4.4 billion of junior debt depended more directly on Anthropic's own credit. The junior debt carried a yield of approximately 8.5 percent, versus a 5.75 percent rate on the senior debt, according to the FT, a 2.75 percentage point spread reflecting the tranche's distinct terms and credit protections.
Questions about that unguaranteed slice predated the announcement. On June 2, Semafor reported, citing people familiar with the matter, that some prospective buyers of the unguaranteed debt had declined to participate because they had not received detailed financial information about Anthropic itself.
What KBRA Had to Work With
The disclosure also reveals, indirectly, how little standard financial disclosure exists for private AI companies. According to the FT, Anthropic supplied KBRA with revenue, adjusted operating income, first-quarter annualized revenue, and internal second-quarter projections. The materials did not include free cash flow or earnings before interest, taxes, depreciation, and amortization.
Denied those conventional measures, KBRA analysts estimated the missing figures and used a risk-scoring approach rather than a full financial model, the newspaper reported. In effect, one of the most consequential credit assessments in the AI sector was built partly on projections and estimation, a sign of how opaque the finances of frontier labs remain even to the agencies grading them.
A Balance Sheet With Almost No Debt
One nuance matters for reading the rating: the FT reported that Anthropic had no debt on its own balance sheet in late June. The company's infrastructure obligations sit with counterparties and special-purpose financing vehicles rather than with Anthropic directly, a structure that separates the lab's direct borrowing from the compute purchases that power it.
That architecture is becoming the industry norm, and it complicates credit analysis. A rating on the operating company says less about the solvency of the financing vehicles its partners construct around it, and vice versa. Investors parsing AI credit risk now have to evaluate multiple layers of the stack, each with different claims on the same future revenue.
Why the Timing Matters
The disclosure lands as banks assemble another wave of financing. On October 2, Bloomberg reported that banks were working on roughly $60 billion in additional chip financing, including a proposed $18 billion junior tranche. Junior tranches, by definition, are the slices most exposed if a borrower's commercial performance disappoints, which makes a formal credit benchmark newly relevant.
Anthropic's financing trajectory has been one of the fastest in the industry, and its revenue growth has been substantial by the company's own disclosures. But the BB+ rating is a reminder that outside evaluators price AI credit risk on the evidence available, and that the evidence, even for a top-tier lab, currently includes neither free cash flow nor EBITDA. As the next $60 billion moves toward the sector, lenders will be grading on a curve that is only now becoming visible.
A Benchmark the Sector Has Lacked
Until Friday, none of the major private AI labs had a publicly known rating from a recognized agency, which meant the hundreds of billions flowing into AI compute deals were being underwritten largely on private diligence. The Anthropic documents change that in one respect: they show how an agency actually treats a frontier lab as a credit, including which disclosures it demands and which it cannot get.
That precedent matters beyond one company. Rating agencies, bank credit committees, and fixed-income investors now have a reference point for the question that has hovered over AI infrastructure financing since the spending boom began: what does a top-tier AI developer look like as a borrower? The answer, at least from KBRA, is speculative-grade and assessed partly on management projections rather than audited cash flows, with a stable outlook attached.
It also arrives during a broader investor debate about how AI compute deals are structured, with billions in vendor arrangements, guarantees, and special-purpose vehicles circulating among a small group of counterparties. A formal rating, even a junk-tier one, gives that debate something it has mostly lacked: an outside, documented data point. Whether other labs follow Anthropic into the rated universe may depend on how the next $60 billion in financing is received.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
