Anthropic is set to tap a $36 billion private credit deal led by Blackstone and Apollo to finance Google AI chips backed by Broadcom, in one of history’s largest debt financings.
Summary
- Apollo and Blackstone are syndicating about $36 billion in debt to finance Google TPUs for Anthropic
- Broadcom will backstop payments as Anthropic leases the custom chips to fuel its Claude AI models
- The deal comes as Anthropic’s valuation jumps to $965 billion and its AI revenue run rate accelerates
Private equity giants Apollo Global Management and Blackstone are working to syndicate roughly $36 billion in debt financing that will fund Anthropic’s next wave of AI infrastructure, in what could become one of the largest private credit transactions ever executed. Reuters reported that the debt will be used to buy custom tensor processing unit chips from Google, which Anthropic will then lease as it races to scale compute for its Claude chatbot and related models.
The structure is straightforward but aggressive: Apollo and Blackstone are inviting additional investors into the $36 billion deal while planning to retain “significant portions” of the debt on their own books, according to additional report by Bloomberg, said that Broadcom, which co designs Google’s TPUs, is backstopping payments on the largest tranches, effectively using its balance sheet and its existing role in Google’s chip stack to de risk the financing. Investors have been asked to submit orders this week, with the transaction expected to close as early as next week, though terms could still shift as books are finalized.
Massive private credit bet on AI compute
For Anthropic, the deal deepens an already sprawling web of capital commitments around AI infrastructure, stacking financial leverage on top of earlier equity deals and long term capacity contracts with hyperscalers. In April, Anthropic secured access to around 3.5 gigawatts of TPU compute through an expanded partnership with Google and Broadcom, with deployments slated to begin scaling from 2027 as part of a broader $50 billion push into domestic US compute capacity. The company simultaneously announced it had raised $6.5 billion in new funding at a $965 billion post money valuation, overtaking OpenAI on paper as demand for its Claude models accelerates.
Anthropic’s revenue run rate recently blew past $30 billion annually, more than tripling from roughly $9 billion at the end of 2025 as its share of the enterprise API market climbed from 12 percent in 2023 to 32 percent by mid 2025, our reporting has showed. That growth has been fueled in part by large financial and industrial clients integrating Claude into production workflows, even as regulators and officials scrutinize emerging risks: in April, US Treasury officials convened major bank CEOs over cyber risks linked to Anthropic’s forthcoming Claude Mythos model after internal tests and a code leak highlighted its ability to uncover “unprecedented” volumes of software vulnerabilities.
Broadcom’s role and the chip supply squeeze
Broadcom’s decision to guarantee payments on large slices of the Anthropic chip financing underscores how AI hardware suppliers are starting to act more like structured finance counterparties than mere vendors. The semiconductor firm already sits at the core of Google’s TPU roadmap and is expected to support future iterations of the chips that Anthropic will be leasing, including next generation designs Google is exploring with partners such as Marvell to improve memory bandwidth and model efficiency.
The Blackstone Apollo transaction is only the latest sign that private equity has no intention of staying on the sidelines of AI infrastructure. Earlier this month Anthropic and a group including Blackstone, Goldman Sachs, Apollo and Hellman & Friedman launched a $1.5 billion venture aimed at pushing Claude into portfolio companies across sectors from healthcare to manufacturing, according to CNBC. For crypto markets, the move reinforces a familiar pattern: capital is clustering around a tiny number of AI platforms that already absorb an outsize share of cloud, chip and power budgets, even as tokenized AI projects scramble for scraps on chain.
