Analysis
Anthropic is reportedly negotiating a $60 billion chip-financing package, Project North Star, that would price junior debt above 10% -- up from the 8.5% its own Project Big Sky financing carried earlier this year, TokenPost reports. Pulse has covered the deal's mechanics in full elsewhere in this issue. The angle that matters more for LPs than the deal itself: a lot of people who think they've diversified away from AI-equity risk haven't.
Private-credit and infrastructure-debt funds are a standard allocation for endowments and fund-of-funds precisely because they're supposed to behave differently from venture equity -- steadier income, different risk drivers, uncorrelated drawdowns. But when the underlying collateral is AI chip-lease payments, that diversification is an illusion. An LP holding a venture fund with AI-lab exposure and a private-credit fund now underwriting that same lab's debt is making the same bet twice: that AI compute demand holds up over a multi-year horizon. If it doesn't, both sleeves get hit by the same event, just through different legal structures and on different timelines.
“But when the underlying collateral is AI chip-lease payments, that diversification is an illusion.”
We've written before about CoreWeave's 9.625% notes touching a roughly 15% yield as bondholders priced in doubt about its $52 billion in combined obligations. That's the same mechanism playing out one rung up the risk ladder: credit investors, not equity investors, are the ones currently re-marking AI infrastructure risk in real time, because debt reprices continuously while private equity marks update quarterly or annually. An LP's venture fund statement won't show this stress for months. The credit fund's statement will show it now.
Room for disagreement: most LPs' credit and infrastructure allocations aren't concentrated in AI-exposed names, and a handful of chip-financing deals repricing upward doesn't mean every private-credit fund is secretly an AI bet. The read-through here is narrower than it sounds -- it applies specifically to LPs who've allocated to data-center, chip-lease or AI-infrastructure-focused credit vehicles, not to private credit broadly.
Still, for the subset of LPs who have done exactly that, the math is simple: two supposedly uncorrelated sleeves, one underlying risk, and a credit market that's already telling you the price of that risk is going up.