What happened
Bank of America, Citi, Morgan Stanley and private‑equity firm Blackstone have jointly structured a $63 billion financing package to fund the acquisition of Broadcom XPU accelerators and networking equipment for Anthropic, an advanced AI model developer. The deal uses a newly created special‑purpose vehicle (SPV) that will purchase the hardware from Broadcom and then lease it to Anthropic. Bank of America, Citi and Morgan Stanley are arranging the senior (first‑lien) portion of the loan, while Blackstone is providing the committed second‑lien tranche. Broadcom will also supply residual‑value support equal to the first‑lien amount, helping to mitigate equipment‑value risk. Deliveries of the Broadcom hardware are expected to begin in calendar year 2027. The transaction is backed by legal advisers Kirkland & Ellis, Latham & Watkins, Sullivan & Cromwell and Evercore, but the parties have not disclosed the loan’s maturity, interest rates, or specific lease terms.
The financing agreement was announced by the four financial institutions and Broadcom in a joint statement on beinsure.com. The SPV will acquire Broadcom’s XPU AI accelerators and associated networking solutions, then lease the equipment to Anthropic for its needs. The first‑lien portion of the loan, arranged by Bank of America, Citi and Morgan Stanley, represents senior secured debt, while Blackstone’s second‑lien commitment sits behind it in the repayment hierarchy.
Broadcom’s role extends beyond supplying hardware; the company will provide residual‑value support equal to the senior loan amount, effectively guaranteeing a portion of the equipment’s future market value. This arrangement is intended to reduce the risk to lenders and align Broadcom’s interests with the success of Anthropic’s deployment.
Legal counsel for the parties includes Kirkland & Ellis (Blackstone), Latham & Watkins (the three banks), Sullivan & Cromwell (Broadcom) and Evercore (financial adviser to Broadcom). The parties have not disclosed the exact allocation of the $63 billion among the lenders, nor have they released details on interest rates, loan maturity, or the lease payment schedule. The first shipments of Broadcom hardware are scheduled for 2027, indicating a multi‑year rollout plan.
Why it matters
The financing marks one of the largest single‑purpose capital commitments ever made for AI infrastructure, underscoring the scale of capital required to train next‑generation foundation models. By separating equipment ownership from usage, the SPV model reduces upfront capital outlay for AI developers like Anthropic, potentially accelerating model development and deployment. The involvement of major banks and a private‑equity firm signals that institutional capital is increasingly willing to back high‑risk, high‑cost AI projects, which could set a precedent for future financing structures in the sector. Moreover, Broadcom’s residual‑value commitment ties semiconductor supply directly to financing terms, highlighting a deeper integration between hardware vendors and financial markets. This arrangement may influence how other AI firms secure compute resources, especially as demand for specialized accelerators and high‑speed networking outpaces traditional corporate borrowing mechanisms.
The $63 billion figure dwarfs most prior AI‑related financing deals, illustrating the massive capital intensity of training large foundation models that require thousands of specialized accelerators and high‑bandwidth interconnects.
By using an SPV to own the hardware, Anthropic can avoid a large upfront CAPEX hit, preserving cash for research, talent acquisition, and other operational needs. This financing model could become a template for other AI startups that lack the balance sheet strength to purchase outright.
The involvement of major banks signals confidence in the long‑term profitability of AI services, potentially unlocking a new asset class for institutional investors. It also reflects a shift toward more sophisticated financing structures that account for the rapid depreciation and obsolescence risk inherent in AI hardware.
Broadcom’s residual‑value commitment ties semiconductor supply directly to financing terms, which may encourage other chipmakers to offer similar guarantees, thereby reducing lender risk and possibly lowering financing costs for AI developers.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
Which component of an AI application is the machine-learning model itself?
What to watch next
Key uncertainties include the specific lease rates, repayment schedule and how the SPV will handle equipment depreciation if AI workloads evolve faster than anticipated. Observers should monitor whether Anthropic secures additional financing for subsequent phases beyond the 2027 delivery window, and whether other AI developers adopt similar SPV‑based leasing structures. The deal also raises questions about regulatory scrutiny of large‑scale AI financing, especially given the involvement of multiple major banks; any changes in banking regulations or capital‑adequacy rules could affect the viability of comparable future transactions.
The specific terms of the lease—such as payment cadence, usage metrics, and end‑of‑lease equipment disposition—will determine the overall cost of for Anthropic and could influence the pricing of AI services built on this infrastructure.
Regulatory bodies may scrutinize the concentration of financing risk among a handful of large banks, especially if similar deals proliferate across the AI sector. Any changes in banking regulations or capital‑adequacy requirements could affect the feasibility of future SPV‑based financing.
Anthropic’s ability to scale its needs beyond the initial 2027 deliveries will be a key indicator of the deal’s success. Follow‑up announcements about additional financing rounds or equipment upgrades will reveal whether the SPV model can accommodate the rapid evolution of AI workloads.