Google has assembled a network of contracts worth more than $150bn to supply Anthropic with chips for its Claude AI models, involving Broadcom, Apollo, Blackstone, Morgan Stanley, and crypto-mining firms turned compute providers. Because Anthropic lacks a credit rating, the deal splits risk: Google sells custom tensor processing units co-designed with Broadcom and guarantees the data centres, while Broadcom commits to buying back unsold capacity and backs a deficiency guarantee covering roughly $31bn of senior debt. Apollo and Blackstone provide private credit through a special purpose vehicle that leases hardware back to Anthropic, with Morgan Stanley advising and lending. The first tranche, a $35bn SPV agreed in June, was the largest private credit transaction on record, priced at around 5.75% due to Broadcom’s guarantee lifting it to investment grade. This vendor financing echoes late-1990s schemes by Lucent and Nortel but is unprecedented in scale, with exposure kept off most balance sheets, including Anthropic’s, ahead of a likely listing. Alphabet owns roughly 14% of Anthropic, making Google supplier, guarantor, and shareholder. Broadcom said its financing platform could support more than 20 gigawatts of compute through 2028 for developers including Anthropic and OpenAI, and rival structures are forming, with OpenAI in talks with Nvidia.