Sources: Apollo Global and Blackstone are working to bring additional investors into a ~$36B debt financing deal to purchase Google TPUs for Anthropic to lease
Context & Ripple Effects
Anthropic's compute relationship with Google had already been discussed as a high-tens-of-billions-dollar arrangement, and Google later described a larger cash-and-performance-linked investment commitment to Anthropic. The reported financing would add a separate capital-markets layer to that compute buildout.
Related reporting indicates Apollo and Blackstone ultimately assembled a roughly $35B package, with Broadcom backing payments on the largest senior debt tranches. That structure matters because it divides the financing into differently priced risk layers rather than leaving all exposure with the lenders arranging it.
First-order effects
- Apollo Global and Blackstone would syndicate part of the proposed TPU acquisition financing to additional investors, broadening the capital base behind Anthropic's leased compute capacity.
- Anthropic gains a route to use Google TPUs without directly funding the full hardware purchase, while Google gains a financed deployment path for its TPU infrastructure.
Second-order effects
- Broadcom's reported support for senior portions can make those tranches more financeable and leave investors in junior or riskier portions to demand higher returns, as reflected in the related coverage's differing yield expectations.
- The arrangement gives private-credit investors a larger role in AI infrastructure funding, while cloud and chip partners become more important in allocating and underwriting the underlying equipment risk.
Third-order effects
- If similar structures recur, AI-compute expansion could increasingly be funded through asset-backed, layered debt arrangements rather than solely through hyperscaler balance sheets or AI-company equity raises.
- That would make the durability of AI infrastructure economics—hardware usefulness, customer payment capacity, and residual equipment value—a central issue for credit markets, not just for model developers and cloud providers.
The trend: This is part of the shift toward structured private-credit financing for the capital-intensive AI compute supply chain.