Four of Wall Street's biggest financial institutions have committed to a $63 billion financing structure that will buy Broadcom's AI chips and networking equipment and lease them to Anthropic, in what the partners describe as a landmark arrangement for funding AI compute capacity at scale.
Bank of America, Blackstone, Citi and Morgan Stanley announced the transaction on Monday, October 26. According to the announcement, a newly formed special purpose vehicle (SPV) with committed financing from the four partners will purchase Broadcom XPU AI accelerators and networking solutions, then lease them to Anthropic. Deliveries are expected to begin in calendar 2027 as part of a previously disclosed strategic compute capacity agreement between Broadcom and the San Francisco-based AI lab. For more context on this story, see our ongoing AI trends.
The deal ranks among the largest private financing packages assembled for AI infrastructure to date, and it lands at the center of an intensifying Wall Street debate over how the industry's enormous compute buildout will be paid for.
How the $63 Billion Deal Is Structured
The financing is split into tranches with distinct risk profiles. Bank of America, Citi and Morgan Stanley will arrange the first lien portion of the debt, while Blackstone serves as the sole provider of the committed portion of the second lien financing — a junior position that carries more risk if the borrower runs into trouble.
Broadcom, for its part, will provide residual value support to the SPV in an amount equivalent to the first lien portion of the financing. That effectively means the chipmaker is standing behind the value of its own hardware, a structure designed to reassure lenders who are being asked to commit tens of billions of dollars against AI equipment with an uncertain resale market.
Bloomberg reported that banks are eyeing around $21 billion of bonds as part of the financing vehicle, marking the beginning of a securitization market for AI chip deals.
Data Center Dynamics reported that the hardware at the center of the transaction comprises Broadcom-developed Google Tensor Processing Units (TPUs) — the custom accelerators Google co-designs with Broadcom and which Anthropic already uses across its data centers.
Hock Tan Personally Intervened to Seal the Deal
The smooth announcement masks a turbulent road to the finish. According to a Wall Street Journal exclusive published on Tuesday, banks initially got cold feet about the scale of the commitment, and Broadcom CEO Hock Tan stepped in personally to salvage the agreement — internally codenamed Project North Star, according to reports.
Tan's direct involvement underscores how much is riding on the deal for Broadcom, which has staked its future on custom AI silicon. The company's XPU business, built on co-designing accelerators with hyperscale customers, has become one of the fastest-growing segments in semiconductors.
The financing is also a coup for Blackstone, the world's largest alternative asset manager, which has been courting AI infrastructure deals aggressively. Jon Gray, Blackstone's President and COO, framed the deal as a template for the asset class: "The rapidly growing demand for AI is creating an enormous opportunity to finance compute capacity at scale. We're proud to help advance this buildout with Broadcom."
Tan echoed that sentiment in the announcement: "Today's transaction underscores the power of our advanced technology and our strong conviction in the AI ecosystem. Fundamentally, the need for AI compute continues to grow at an extraordinary pace."
A Rapidly Growing Web of AI Debt
The finalized $63 billion package is the latest step in a fast-moving financing relationship between Broadcom and Anthropic. Earlier this year, reporting on Anthropic's IPO prospectus revealed that Broadcom planned to lend the AI lab up to $42 billion to support chip leasing, and August reporting suggested Broadcom's ultimate AI debt ambitions for Anthropic could climb as high as $100 billion.
The structure also echoes a broader pattern in which lenders, chipmakers and AI labs are binding themselves together in circular arrangements: chipmakers finance the purchases of their own hardware, banks securitize the debt, and AI labs lease compute they could not otherwise afford to buy outright.
That pattern has not gone unquestioned. Market commentators — including CNBC's Jim Cramer — have warned investors about mistaking financing-driven AI stock rallies for durable demand, and analysts note that deals like Project North Star deepen Wall Street's exposure to AI infrastructure risk at a moment when questions about AI revenue growth remain unresolved.
Perfect Timing for Anthropic's IPO
The announcement landed on the same day Anthropic officially filed for its long-awaited initial public offering. The company plans to list on Nasdaq under the ticker ANT, capping a year in which it reported multi-billion dollar revenue figures and secured a series of enormous compute agreements across chipmakers, cloud providers and data center operators.
For Anthropic, the SPV structure means access to tens of billions of dollars of compute capacity without an equivalent upfront cash outlay — critical for a company that must balance IPO-era financial disclosures against the compute demands of training and serving frontier models.
For Broadcom, the deal validates its position as the indispensable merchant supplier of custom AI silicon, and it gives the company a template it can replicate with other hyperscale and frontier customers.
What Comes Next
The financing vehicle's first bond sales are expected to be structured in the coming months, with Bloomberg reporting banks are targeting roughly $21 billion in an initial offering. Deliveries of Broadcom accelerators and networking gear to Anthropic begin in 2027, giving both companies a multi-year revenue and deployment runway.
Whether Wall Street's appetite for AI paper holds at this scale is the open question. For now, the four institutions behind Project North Star have made their bet — and it is one of the biggest wagers yet on the economics of frontier AI.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
