A single document has put a price on the AI boom’s most tangled partnership. Anthropic’s IPO prospectus, first reported by Reuters, discloses that Broadcom will extend the Claude developer as much as $42B in credit.
The scale matters to both sides. Anthropic has pledged $125.2B over five years to lease tensor processing units, and Broadcom’s lending would carry close to a third of that. Because the two firms co-design those chips with Google, Broadcom is effectively helping pay for hardware it also supplies: capital leaves a chipmaker, turns into compute at a lab, and comes back as orders.
Structured as convertible notes, the facility could hand Broadcom equity in Anthropic. Regulators were told the company has no plan to sell any notes ahead of the listing’s completion.
Analysts and watchdogs see a bigger pattern. The Bank for International Settlements, in a report issued Thursday, put 46.4% of AI-to-AI deal value across 2021 to 2025 in transactions between firms with commercial supply-chain links. Rothschild & Co managing partner Robert Leitao described the Broadcom arrangement to Reuters as a concentrated wager that two companies can earn enough to service all the financing stacked behind them.
Broadcom has been here before: in June it joined Apollo and Blackstone to stand up an AI infrastructure financing platform. That vehicle’s opening transaction, worth $35B, backed more than a gigawatt of Anthropic capacity. By 2027, the projection runs the other way, with Anthropic expected to stand as Broadcom’s single largest buyer of compute.
