Bank of America, Blackstone, Citi and Morgan Stanley will provide $63 billion in financing to buy Broadcom AI chips and networking equipment for Anthropic, in a deal announced on Monday, October 26, 2026, that shows how Wall Street, not just venture capital, is now funding the AI buildout. A newly formed special purpose vehicle will use the money to buy Broadcom’s XPU accelerators and related gear, then lease that hardware to Anthropic, with deliveries expected to begin in 2027. Broadcom will backstop the senior lenders by providing residual value support equal to the first-lien debt, a promise to absorb declines in the equipment’s value up to that amount.
How The $63 Billion Is Structured
The structure matters because Anthropic is not borrowing $63 billion in cash to spend as it pleases. The vehicle borrows, buys specific Broadcom hardware, and leases it to the company behind Claude. Bank of America, Citi and Morgan Stanley are arranging the first-lien financing, which ranks first for repayment if the vehicle runs into trouble. Blackstone is the sole provider of the committed second-lien financing, which ranks below the senior debt and normally carries a higher return for that extra risk.
The companies did not disclose the split between the two layers in their announcement. Reuters, cited in reporting on the deal, put the first-lien portion at about $43.5 billion, or roughly 69 percent of the total, based on people familiar with the matter. That figure is worth treating as reported, not confirmed, because the official release leaves the breakdown out. What the release does confirm is the shape of Broadcom’s support: residual value support for an amount equal to the first-lien portion, designed to reassure senior lenders that a fall in chip prices will not leave them holding the loss alone.
Broadcom chief executive Hock Tan said the deal validates that capital is available to finance AI growth at scale. Blackstone president Jon Gray called the demand for AI an enormous opportunity to finance compute capacity, while Morgan Stanley co-president Dan Simkowitz called the deal an important milestone in the next buildout of AI compute capacity.
Not The First Lease, And Not A Small One
This is the second large lease-backed structure of its kind for Anthropic. An earlier deal, known as Big Sky, raised $35 billion, most of it supported by Broadcom, and received an investment-grade rating. The new $63 billion transaction is therefore not an isolated experiment. It extends a model in which a chipmaker helps finance its own customer’s purchases, with banks and asset managers supplying the debt and the chipmaker absorbing part of the residual-value risk.
The timing is also telling. Anthropic filed publicly for an initial public offering on the same day, October 26, the first listing by a company focused primarily on developing frontier AI models. The filing gave investors a much clearer look at a balance sheet built for enormous spending. As of June 30, Anthropic reported $6.2 billion of current debt in the form of convertible promissory notes to Amazon, plus a five-year $15 billion revolving credit facility led by Morgan Stanley that had been increased from about $3 billion in September.
The filing also showed leases not yet commenced jumping to about $321 billion at the end of June, up from about $50 billion in December, and an option to borrow another $77 billion in convertible notes from Amazon, Broadcom and Google. Those numbers explain why an investment-grade rating matters: it would open a deeper pool of capital and cheaper borrowing. So far, the main rating agencies have yet to publish ratings for Anthropic.
Why Lenders Need A Backstop
AI accelerators are not like office buildings. They lose value fast as newer chips arrive, which makes lenders nervous about financing them over several years. That is why the residual value support is the most important clause in this deal, and why similar structures are drawing scrutiny across the industry. The same week, analysts were warning about the debt behind the wider AI buildout, as covered in our report on Moody’s warning over AI’s $1 trillion buildout, and SpaceX closed a comparable lease structure of its own, as in our report on SpaceX’s $40 billion Apollo chip lease. The pattern is consistent: the hardware is too expensive to buy outright, too fast-depreciating to finance without a backstop, and too strategically important for the chipmaker to let the deal fail.
Why It Matters
This matters because it shows who is really paying for AI’s next phase. The first wave was funded by venture capital and big tech balance sheets. The next wave is being funded by banks, asset managers and chipmakers themselves, using debt structures that tie their fates together. If Anthropic’s models generate the revenue to pay the rent, the structure looks like a clever way to scale fast without diluting shareholders. If demand disappoints, or if chip values fall faster than expected, Broadcom’s backstop and the banks’ senior position will be tested in ways that a simple equity round never would be.
Background
Anthropic, the maker of the Claude chatbot, filed publicly for an IPO on October 26, 2026, with Goldman Sachs and Morgan Stanley leading the offering. Its earlier Big Sky financing, backed by Broadcom and financed with Apollo and Blackstone, raised $35 billion for AI computing capacity.
Source: Business Times report
0 Comments