Broadcom to lend Anthropic up to $42B, most of it paid back to Broadcom

Broadcom is on both sides of the deal

On October 1, Reuters obtained Anthropic's IPO prospectus and reported that Broadcom has agreed to lend the company up to $42 billion to fund its infrastructure buildout. The part that turns a routine-sounding credit facility into a structural story, according to a dev.to analysis of the filings, is the destination of the money: Anthropic borrows from Broadcom and then spends most of it leasing compute hardware from Broadcom.

The facility could cover roughly one-third of a $125.2 billion, five-year TPU lease commitment disclosed in the same IPO filing, and Broadcom is already installing one gigawatt of Ironwood chips for Anthropic this year. The commitment also stacks on an earlier $35 billion tranche arranged in June through a financing platform Broadcom set up with Apollo Global Management and Blackstone — a structure Broadcom then described as involving no direct lending, only "modest residual value guarantees." The dev.to analysis reads the new $42 billion facility as a substantially deeper entanglement than that June framing implied.

There is a covenant clause worth noting. Anthropic funded a restricted account for Broadcom's benefit in April and may have to add to it. Should Anthropic miss a payment or breach another condition, a large share of its lease obligations could accelerate and become due at once — exactly when access to the $42 billion facility meant to service them is restricted. Broadcom ends up as both lender and lessor, holding first claim if its customer runs into trouble.

Not an isolated deal

The report places the Broadcom-Anthropic arrangement inside a two-year pattern across the AI stack. Nvidia has structured similar arrangements with OpenAI, including up to $100 billion in direct investment tied to 10 gigawatts of data center capacity, plus reported discussions around a $250 billion guarantee and $350 billion in chip-purchase financing for a single Ohio campus. Oracle has committed roughly $300 billion in purchases tied to OpenAI, Microsoft around $250 billion, and OpenAI itself has committed $90 billion to AMD while taking an equity stake in Nvidia.

Investor Michael Burry, known for shorting subprime mortgages before 2008, has publicized a Bloomberg-sourced diagram tracing roughly $46 billion in direct equity stakes and $879 billion in multi-year purchase commitments among Microsoft, Oracle, Amazon, Google, Meta, OpenAI, Anthropic, xAI, CoreWeave, Nvidia and AMD. His argument, as summarized by dev.to: when a hyperscaler funds a lab and the lab spends that money buying compute from the same hyperscaler, the spend is booked as revenue, and after enough loops the top line stops telling you much about real customer demand.

The Bank for International Settlements adds harder numbers, estimating that the five largest hyperscalers carry roughly $1.65 trillion in off-balance-sheet debt through special-purpose vehicles — more than the $1.35 trillion they report directly. That exposure surfaces in credit markets rather than standard leverage ratios, and dev.to notes that Nvidia's five-year credit default swap spread has roughly doubled over the past two months.

The counterargument

The circular-financing framing has prominent critics. Nvidia chief executive Jensen Huang has called it "ridiculous," and the company generated close to $48 billion in free cash flow in a recent quarter — real cash rather than an accounting artifact. CoreWeave CEO Michael Intrator has pointed out that Nvidia's roughly $300 million stake in his company is small next to the more than $25 billion CoreWeave raised independently. Wedbush analyst Dan Ives frames the current moment as "1996, not 1999": early-cycle adoption rather than late-stage mania, with enterprise AI uptake still a single-digit share of addressable spend.

Both readings can hold at once, the dev.to analysis argues. Vendor financing is a normal feature of capital-intensive industries building ahead of demand — telecom did it in the late 1990s, and some of that fiber is still in use — so the technology can be genuinely valuable while the financing around it stays fragile. The specific problem is concentration and circularity: when the same handful of counterparties are simultaneously each other's lender, customer, supplier and shareholder, a stumble at any one node propagates quickly.

Why it matters

For engineers and founders building on top of these APIs, the financing web is a dependency risk that rarely gets priced in. Heavily leveraged, interconnected providers have strong incentives to protect margins if credit conditions tighten — through price increases, rate limits or deprioritizing less profitable workloads. Multi-vendor fallbacks, usually treated as an optional resilience pattern, start to look like basic financial risk management given how intertwined the major labs now are with each other's funding. And with Anthropic and OpenAI combined at roughly $1.8 trillion in valuation while unprofitable, any repricing of AI-linked debt would likely hit API pricing and feature availability long before it shows up on a balance sheet. None of this means the models stop working next quarter; it means the companies behind them make pricing and roadmap decisions under leverage, which is worth factoring into any assumption that today's API terms stay stable for the next three years.