Anthropic offers Blackstone more than 10% to finance chips as AI borrowing costs climb

Anthropic is offering Blackstone a yield of more than 10% in its latest chip-financing deal, Reuters reported, citing the Financial Times^1,2. The FT set the pricing against the ICE BofA BB US High Yield Index and reported that the debt could become tradeable in the institutional 144A bond market¹.

The pricing arrives with the rest of the AI credit complex under pressure. The New York Times reported that investors have sold off bonds for other AI giants as well, including Larry Ellison’s Oracle, though those price declines are not as steep⁵. Semafor reported that Oracle’s $160 billion of debt is rated just above junk status and that insuring Microsoft’s AAA-rated debt costs 25% more than it did three months ago⁷. Reuters reported that CoreWeave’s debt stress is testing private credit’s AI financing boom, with James Martin, senior US credit strategist at SMBC, saying investors were favouring high-yield data-centre financings for projects closer to completion⁸.

Why it matters to investors

A double-digit private yield is a price, and prices in this market have been scarce. The Dark Side of the Boom noted that Goldman tracks a universe of 20 dollar high-yield project finance bonds with around $45.6 billion of total notional outstanding³, a thin sample set against the scale of the buildout. The same writer framed Anthropic’s $500 billion valuation as a reality check on what happens when the AI growth curve meets the cost of capital⁴.

The structural question is who ends up holding the paper. If the debt reaches the 144A market it becomes tradeable, and the marks become public¹. Until then the financing sits with private lenders. The FT, citing the Bank for International Settlements, warned that we must not assume this will go on forever and must be ready for systemic shocks when the music stops⁶.

What to watch

Watch whether the chip financing is syndicated into the 144A market, because that is the moment a private yield becomes a public mark^1,2. Watch the cost of insuring the investment-grade hyperscalers, which Semafor’s Microsoft figure shows is already moving⁷. And watch whether lenders keep retreating toward near-complete projects, the pattern SMBC described, since that is how a funding squeeze reaches the earliest-stage data centres first^5,8.