Sources: a Broadcom-backed portion of Apollo and Blackstone's $36B debt deal to buy TPUs for Anthropic to lease may yield ~5.75% vs. 8%-9% for a riskier portion
Broadcom Inc.'s promise to backstop the largest portion of a record-setting microchip financing deal for Anthropic is keeping yields …
Context & Ripple Effects
The financing was preceded by Apollo and Blackstone’s talks with Broadcom over funding AI-chip development, then efforts to add investors to a roughly $36 billion TPU transaction. Separate coverage had also tied Anthropic to large orders of Google TPU racks, establishing the compute-demand backdrop for the debt package.
Subsequent related reports say the package was finalized at about $35 billion and that Broadcom would backstop payments on its largest senior portions. The reported yield split shows how that support is being translated into the deal’s capital structure.
First-order effects
- The Broadcom-backed senior debt can be offered at a materially lower yield than the riskier portion, reducing the financing cost for the most protected part of the TPU acquisition.
- Apollo, Blackstone, and other lenders gain a clearer division between exposure supported by Broadcom and residual risk tied more directly to the lease arrangement and underlying assets.
Second-order effects
- Investor demand is likely to concentrate in the supported senior tranches, while junior or otherwise riskier debt must pay more to clear the market; the financing’s economics therefore depend heavily on the scope of Broadcom’s backstop.
- Chip vendors and AI customers pursuing similar infrastructure financings will face pressure to provide stronger contractual support if they want private-credit capital priced closer to senior secured risk.
Third-order effects
- If replicated, large AI-compute deployments could increasingly be financed through layered private-credit structures in which supplier guarantees or backstops determine pricing, rather than through a single undifferentiated loan.
- That model may deepen the interdependence of AI developers, hardware suppliers, and alternative-asset managers, while concentrating financial importance in the vendors able to support customer payment obligations.
The trend: AI compute is becoming a structured-finance asset class, with credit protection from major hardware suppliers separating low-cost senior funding from higher-yield residual risk.