On October 26, 2026, Bank of America, Blackstone, Citi and Morgan Stanley announced a $63 billion financing structure to pay for Broadcom AI accelerators and networking gear destined for Anthropic. The number is large enough to grab attention. The mechanics underneath it are the more interesting story, because they show how the most capital-hungry software companies in the world are now buying compute: not with equity alone, but through leases, tranched debt and guarantees from the chip supplier.
How the structure works
According to reporting by Reuters and Yahoo Finance, a newly formed special purpose vehicle will purchase Broadcom’s XPU AI accelerators and networking products and lease them to Anthropic. Bank of America, Citi and Morgan Stanley are arranging the first-lien debt, while Blackstone is the sole provider of the committed second-lien portion. Deliveries are expected to start in 2027 under a compute capacity agreement Broadcom had disclosed earlier.
The companies did not publish a split between the layers. Two sources told Reuters that the senior first-lien loans come to roughly $43.5 billion, about 69% of the total. If that is right, the second-lien slice would be around $19.5 billion, though that figure is our arithmetic and not a disclosed number.
The part that holds it together: residual value support
The key ingredient is a backstop. Broadcom is providing what the announcement calls residual value support to the vehicle, in an amount equivalent to the first-lien financing. In plain terms, if the hardware turns out to be worth less than expected when the lease ends, or the lessee cannot pay, the chip maker absorbs part of that loss. Reuters notes that chipmakers have offered similar guarantees before to protect lenders against falling values of advanced chips and data center assets. In an earlier arrangement covering a $35 billion expansion of Anthropic’s capacity, financed by Apollo and Blackstone, Broadcom’s backstop was reportedly up to $29 billion, per a Broadcom filing cited by Reuters.
That backstop explains why the lenders were willing. AI accelerators depreciate quickly and are replaced on short cycles, and a lender who finances them wants confidence about what a used chip is worth in five years. By taking on that risk, Broadcom effectively lends its own balance sheet to its customer. Moody’s described Broadcom as one of the strongest large technology companies, but warned that risk would rise if the company completed multiple such transactions in a short period to serve the large funding needs of frontier AI labs.
What the principals said
Blackstone President and COO Jon Gray said that rapidly growing demand for AI is creating an enormous opportunity to finance compute capacity at scale. Broadcom CEO Hock Tan said the need for AI compute continues to grow at an extraordinary pace. Broadcom shares rose about 0.6% in after-hours trading on the day of the announcement, per Yahoo Finance, and AI Weekly’s October 27 roundup reported a gain of roughly 4% in the following session.
One outlet reported that banks had hesitated and that Tan intervened in late talks, reportedly agreeing to higher interest rates and stronger loss protection. That account comes from a single report and we could not corroborate it, so treat it as unconfirmed.
The IPO backdrop
The timing is notable. Reports say Anthropic filed for an initial public offering around the same time, with a planned Nasdaq listing under the ticker ANT. Secondary coverage summarizing the filing cites roughly $16.3 billion of first-half 2026 revenue and an $80.7 billion net loss that is described as largely non-cash, tied to convertible financing instruments, with an operating loss of about $693 million. These numbers come from aggregator summaries of the filing and press accounts, and earlier coverage of the IPO process disagreed on filing status, so verify against the filing itself before relying on them. What can be said is that a company about to ask public investors for money is simultaneously lining up tens of billions in lease financing for its hardware.
Why this matters beyond one deal
For years the AI buildout was framed as a contest between cloud giants spending from cash flow. This deal shows a second channel maturing: private credit and big banks packaging compute into financeable assets, with the chip supplier guaranteeing part of the value. It resembles aircraft leasing, where a lessor owns the expensive asset and rents it to an operator. The risk is also similar. If model efficiency improves faster than expected, or demand softens, the collateral loses value and the guarantor, in this case Broadcom, feels it first.
It also changes who captures margin. Chip designers get a customer that can place very large orders without a matching cash outlay up front. Lenders earn yield from a borrower with a strong revenue trajectory. And the AI lab converts what would be a huge capital expenditure into a recurring lease cost, preserving cash for training and operations. Everyone in the chain gains as long as the assumptions hold.
What This Means for Philippine Founders
First, expect the cost and availability of AI compute to stay tied to financing conditions, not just technology. If lease-backed structures keep scaling capacity at the frontier labs, API capacity should keep growing, but pricing will also reflect the debt service behind it. Philippine startups building on top of these models should avoid architectures that depend on a single provider’s pricing staying flat, and should budget for price changes in either direction. Keeping the option to switch between providers is a practical hedge.
Second, the structure is a reminder that asset-backed financing is becoming normal for infrastructure. Whether similar lease or vendor-backed arrangements will reach smaller buyers in the Philippines, such as local data center operators or GPU cloud startups, is unconfirmed, and nothing in the reporting suggests it will soon. Founders in that niche could still ask suppliers and local lenders about residual-value or vendor-financing terms, since the logic of the deal applies at any scale: the lender needs to believe the hardware retains value.
Share this article