Sources: Apollo and Blackstone finalized a $35B package for Anthropic to lease TPUs; Broadcom is backstopping payments on the debt's largest senior portions
Apollo Global Management Inc. and Blackstone Inc. have finalized a $35 billion financing package for Anthropic PBC to expand its AI infrastructure …
Context & Ripple Effects
This financing moved from talks between Apollo, Blackstone, and Broadcom over funding AI chips to a proposed roughly $36 billion TPU-purchase structure that sought additional investors. The finalized package is slightly smaller at $35 billion and is tied to Anthropic leasing the equipment.
The deal also follows Anthropic's partnership with Blackstone and Hellman & Friedman on Ode, an AI implementation venture. Together, the coverage shows private-capital firms engaging Anthropic both in deploying AI products and in supplying the infrastructure needed to run them.
First-order effects
- Anthropic gains a financed route to expand TPU capacity without directly funding the full equipment purchase itself; Apollo and Blackstone become the lenders and owners behind that leasing structure.
- Broadcom's payment backstop on the largest senior debt portions shifts part of the immediate credit risk away from senior lenders, supporting the package's tiered risk-and-return structure.
Second-order effects
- The backstop can make the senior portion more financeable than the riskier tranche, concentrating lender attention on how much vendor support is available rather than on Anthropic's lease obligations alone.
- The transaction gives private-credit managers a prominent AI-infrastructure asset-financing template, while increasing the importance of hardware suppliers' willingness to support customer financing.
Third-order effects
- If replicated, AI compute expansion could increasingly be funded through structured leases and private credit rather than solely through model developers' balance sheets, tying infrastructure growth more closely to capital-market conditions.
- Vendor-backed senior financing may divide AI infrastructure funding into lower-risk, supported layers and higher-yield residual layers; whether that structure scales depends on sustained equipment utilization and the durability of backstops.
The trend: AI infrastructure is becoming an asset-finance market in which model developers, hardware vendors, and private-credit firms share the cost and risk of securing compute capacity.