Apollo, Blackstone finalize $35 billion AI chip financing for Anthropic

Apollo and Blackstone finalized a $35 billion financing package for Anthropic, one of the largest private credit transactions on record, to fund AI infrastructure tied to Google TPUs. The structure includes $6 billion of A1 notes, $24 billion of A2 notes, and a $4.5 billion B tranche, with Broadcom backing the senior portions via a residual value agreement that helped secure investment-grade private ratings and lower borrowing costs. The deal highlights strong demand for AI infrastructure financing and follows Anthropic's confidential U.S. IPO filing and a new funding round valuing the company at $965 billion.

Analysis

This is not just another AI capex headline; it is a financing template that transfers the first-loss risk on compute from hyperscalers to private credit and structured finance. The key second-order effect is lower marginal cost of funding for AI infrastructure, which should extend the runway for aggressive chip deployment even if operating cash flow lags, making supply-chain winners more durable than the underlying model economics would otherwise justify.

Broadcom is the clearest structural winner because it is effectively monetizing balance-sheet confidence as a product: residual value support turns commodity-like silicon into quasi-asset-backed paper with investment-grade optics. That should widen Broadcom's moat with enterprise and finance buyers, and it may also raise the bar for competing semiconductor vendors that cannot offer similar credit enhancement. For Google, the economic signal is subtler: it validates TPU demand and keeps its custom silicon relevant, but it also locks more compute into a capital structure that is exposed to utilization and resale assumptions over a multi-year horizon.

The main risk is that the market is extrapolating financing capacity faster than actual AI monetization. If model revenue growth or customer retention disappoints over the next 6-18 months, the residual-value support becomes the real stress point, not lease coverage, and the structure could reprice sharply. The broader consequence is that private credit may start absorbing technology risk without being paid enough spread for correlation during an AI capex slowdown, which is exactly when liquidity tends to vanish.

Contrarian read: this is mildly bullish for the AI buildout, but probably more bullish for lenders and silicon than for the platform names already priced for perfection. The market may be underestimating how much this favors structured-credit specialists and capital-light enablers versus the end users of compute, whose return on invested capital still needs to catch up to their funding appetite.

Key Decisions for Investors

• Long AVGO vs a basket of high-multiple AI software names for 3-6 months; thesis is that credit-enhanced compute monetization improves semiconductor cash-flow visibility while software remains vulnerable to valuation compression if monetization slows.
• Buy BX on weakness, but hedge with a short in a broader private-markets proxy over 1-2 quarters; upside comes from origination/placement fees and repeat AI-finance mandates, while downside is a widening of private-credit spreads if the AI capex cycle cools.
• Pair long GOOGL / short NASDAQ-100 growth basket for 1-3 months; if the market keeps rewarding custom silicon and infrastructure optionality, GOOGL benefits from embedded TPU demand without paying the same multiple for pure software duration.
• Consider a tactical short in lower-quality AI beneficiaries with no financing edge if AI sentiment remains risk-on; the trade is that capital markets will increasingly discriminate between firms that can securitize compute and those that merely consume it.