.png "Global Insight")
Wall Street has just placed one of the biggest bets in the history of artificial intelligence. Bank of America, Blackstone, Citi and Morgan Stanley will provide $63 billion in financing to buy Broadcom's AI chips and networking equipment for Anthropic, the company behind the Claude chatbot — one of the largest debt deals ever tied to the AI buildout. The hardware does not even exist yet in full: deliveries are expected to begin in 2027. What exists is a financial structure, a lease, and an enormous amount of confidence that demand for AI computing will keep growing fast enough to pay for it all.
How the deal actually works
The transaction runs through a newly formed special purpose vehicle — a separate company created just for this deal. That vehicle will use the committed financing to buy Broadcom's XPU AI accelerators and networking solutions, and then lease them to Anthropic. Bank of America, Citi and Morgan Stanley are arranging the first-lien financing — the senior debt that gets paid first — while Blackstone is the sole provider of the committed portion of the second-lien financing, which ranks below it. According to Reuters, sources familiar with the matter put the first-lien loans at about $43.5 billion, or 69% of the total, though the official announcement did not disclose the breakdown. Broadcom itself is providing "residual value support" to the senior lenders — effectively agreeing to absorb declines in the value of the equipment up to an amount equal to the first-lien debt. In plain terms: Wall Street lends the money, a middleman company owns the chips, Anthropic rents them, and Broadcom promises the lenders that the chips will hold enough value to cover the senior loans.
Why Anthropic needs this much hardware
Anthropic's models are trained and run on enormous amounts of computing power, and that power has to be bought years in advance. The $63 billion deal is part of a previously announced strategic compute capacity agreement with Broadcom, and it follows an earlier agreement under which Apollo and Blackstone agreed to finance a $35 billion expansion of Anthropic's AI computing capacity. Anthropic's IPO filing has already disclosed the scale of its commitments: Broadcom had agreed to lend Anthropic up to $42 billion to finance infrastructure spending, and the convertible notes in that arrangement could finance about a third of a $125.2 billion commitment Anthropic has made for a five-year lease of tensor processing unit computing capacity. Anthropic filed for a potentially record-breaking IPO on Monday, according to Reuters, even as disclosures of steep losses and security concerns threatened to test investor appetite. The message of the $63 billion deal is that lenders are willing to fund the buildout anyway — for now.
What the cheerleaders say — and what the sceptics say
Blackstone president Jon Gray called the rapidly growing demand for AI "an enormous opportunity to finance compute capacity at scale", while Morgan Stanley co-president Dan Simkowitz called the deal an important milestone in enabling the next buildout of AI compute capacity, according to the companies' joint announcement. Citi's head of banking, Vis Raghavan, said financing AI demands capital solutions at unprecedented scale. The sceptics' version of the same facts is less comfortable: a concentrated bet on two companies — one AI lab and one chipmaker — generating enough revenue to support tens of billions of dollars of financing, with lenders, chipmakers and AI labs all financing each other in a circle. As one fund manager quoted in earlier coverage of the Broadcom relationship put it, there is a concentrated bet on two companies being able to generate enough revenues to support all the financing that has happened. If AI demand keeps compounding, the leases pay for themselves and everyone involved looks prescient. If it stalls, the structure that makes this deal possible — debt, leased hardware, and a chipmaker's value guarantee — is also the structure that transmits the shock. Either way, the $63 billion is a marker of how far Wall Street is now willing to go to own a piece of the AI future without owning an AI company at all.