Broadcom raises $60B to fund AI chips for Anthropic

Broadcom is assembling one of the largest debt financings in the AI chip industry. Banks working for the semiconductor company have begun raising $60 billion to fund purchases of Broadcom's AI chips by Anthropic and other companies, Bloomberg reported on October 2, 2026 (UTC). The deal expands a $42 billion loan to Anthropic disclosed just one day earlier into a far broader financing machine.

The structure is split into two tranches. A $42 billion Class A senior-secured piece — the top-ranking debt, backed by collateral — will be syndicated to investors through Wall Street banks. A separate $18 billion Class B junior tranche is being led by Blackstone, which is committing $9 billion from its own funds and plans to sell the remaining $9 billion to other investors. Broadcom declined to comment.

This is not a new idea for the AI industry. Nvidia announced a similar arrangement in August 2026 (UTC), partnering with six large financial firms — including Blackstone — to raise more than $500 billion for AI infrastructure, including financing to help customers buy Nvidia chips. Broadcom is now building its own version, scaled down but structurally identical: the chipmaker helps finance the customer that buys its products.

How the pieces fit together

The $60 billion package connects to the $42 billion loan that Broadcom agreed to extend to Anthropic, revealed in Anthropic's IPO prospectus and reported by Reuters on October 1, 2026 (UTC). That loan can convert into Anthropic shares and would fund roughly one-third of Anthropic's $125.2 billion five-year lease commitment for TPUs — the custom AI processors that Google designs with Broadcom.

Anthropic is expected to become Broadcom's largest compute customer in 2027, according to Barron's. Broadcom projects AI semiconductor revenue of roughly $115 billion in fiscal 2027 and $230 billion in fiscal 2028. The financing arrangement has been in the works for weeks — people familiar with the talks said in August 2026 (UTC) that it would help Anthropic and other companies obtain chips and AI infrastructure.

Blackstone already holds a stake in Anthropic, an investment that Bloomberg reported has helped boost returns at Blackstone's private-equity fund for wealthy investors. The firm's dual role — equity investor in Anthropic and lead arranger of the junior debt tranche — ties its fortunes even more closely to the AI lab's success.

The circular financing problem

The deal raises a question that has been hovering over the AI infrastructure boom: what happens when the supplier finances the customer?

Under this model, Broadcom lends money so that Anthropic can buy Broadcom chips. Broadcom's revenue depends on Anthropic's ability to pay. If Anthropic stumbles — if revenue falls short, if the IPO is delayed, if AI demand cools — Broadcom is exposed both as a supplier and as a lender. The Anthropic prospectus itself acknowledges that Broadcom's dual roles create potential conflicts of interest.

Robert Leitao, managing partner at Rothschild & Co, put it bluntly: "It feels that there's quite a concentrated bet right now on two companies being able to generate enough revenues to support all the financing that's happened."

The prospectus adds that certain defaults could make a substantial portion of Anthropic's lease obligations immediately due. In a worst-case scenario, a covenant breach could trigger a cascade: lease payments come due, Broadcom's loan is at risk, and the chipmaker's revenue forecast — built partly on Anthropic becoming its largest customer — gets revised downward.

Why it matters

The $60 billion figure is large, but the more important number is the ratio. Broadcom's entire market capitalization is roughly $700 billion. Raising $60 billion in debt to finance customer purchases means the company is leveraging its balance sheet to the tune of nearly 9% of its market value — all to support a single customer ecosystem. Compare that to Nvidia's $500 billion+ partnership, which spreads risk across six financial firms and many more customers. Broadcom's version is more concentrated and, arguably, more fragile.

The deal also reveals how the AI chip war has shifted from a technology competition to a financing competition. Nvidia's advantage is not just GPU architecture — it is the ability to offer customers a complete package of chips, networking, and financing. Broadcom is responding by building its own financing arm, but it is doing so with a customer base that is far narrower. Anthropic is the anchor tenant; without Anthropic's growth, the $60 billion structure has no reason to exist.

There is a timing element worth noting. Anthropic is preparing for an IPO targeting a valuation of $1.8 to $2 trillion, with a pre-IPO investor day scheduled for October 14, 2026 (UTC) in San Francisco and marketing expected to begin the week of November 9, 2026 (UTC). The Broadcom financing gives Anthropic a concrete infrastructure funding plan to present to investors — proof that the company can secure the compute it needs without relying solely on Google's TPU supply. That is a meaningful selling point for the IPO roadshow.

The critical question is whether this financing model is sustainable. If AI revenue grows as projected — Broadcom's $230 billion fiscal 2028 target implies roughly 2x growth from 2027 — the debt is easily serviceable. If growth slows, the circular financing structure becomes a liability rather than an asset. Investors are watching closely amid public opposition to new data-center construction and growing concerns about how much the AI industry is spending.

What to watch next

The syndication letters for the $42 billion Class A tranche will tell us whether institutional investors have appetite for AI chip debt at this scale. If the tranche is oversubscribed, it signals that Wall Street remains bullish on AI infrastructure financing. If it struggles, Broadcom may need to offer higher yields or scale back the package.

Also worth tracking: whether other AI chip designers — AMD, Marvell, or emerging players — launch similar financing programs. The supplier-financed-customer model is spreading, and each new entrant increases the total debt load in the AI supply chain.

One prediction: by the end of 2026, at least one major AI chip financing deal will face a rating downgrade or a pricing reset as credit agencies begin to scrutinize the circular-financing structure more carefully. The model works in a growth environment; it has not been tested in a downturn.