Bank of America, Blackstone, Citi and Morgan Stanley have announced a $63 billion financing transaction to support new artificial intelligence compute infrastructure for Anthropic, using Broadcom’s AI accelerators and networking technology. The deal will be structured through a special purpose vehicle (SPV) that purchases the hardware and leases it to Anthropic, with deliveries expected to begin in 2027. The arrangement highlights how banks and institutional investors are developing large-scale financing structures to fund the capital-intensive infrastructure behind advanced AI systems.

$63 Billion AI Infrastructure Deal Brings Banks and Private Capital Together

Four major financial institutions are backing a $63 billion financing structure designed to expand Anthropic’s artificial intelligence computing capacity, illustrating how investment banking and private capital are becoming central to the infrastructure supporting advanced AI models.

Announced on October 26, 2026, the transaction brings together Bank of America, Blackstone, Citi and Morgan Stanley, with semiconductor and infrastructure technology provider Broadcom supplying the AI accelerators and networking equipment. The hardware will be acquired through a newly established special purpose vehicle and leased to Anthropic, the company behind the Claude AI assistant. Deliveries are expected to begin in 2027.

The arrangement is part of a previously announced strategic compute capacity agreement involving Broadcom. Rather than simply funding software development or purchasing cloud services, the structure finances the underlying computing equipment required to train and operate AI models.

How the $63 billion financing structure works

The transaction separates the financing and ownership of AI hardware from its use by Anthropic. A special purpose vehicle, or SPV, will purchase Broadcom XPU AI accelerators and networking solutions before leasing the equipment to the AI company.

Bank of America, Citi and Morgan Stanley are arranging the first-lien financing, which gives senior lenders priority over junior creditors in claims against the relevant collateral. Blackstone is the sole provider of the committed portion of the second-lien financing, which carries a lower repayment priority.

Broadcom will provide residual-value support to the SPV equivalent to the first-lien financing amount. That provision is designed to address an important consideration in equipment-backed lending: what happens to the value of specialized hardware over the life of the financing.

Reuters reported that approximately $43.5 billion of the transaction will be senior, first-lien loans, representing roughly 69% of the total. The full economic terms, including pricing, repayment schedules and detailed risk allocations, have not been disclosed in the announcement. Reuters+1

The distinction matters because AI accelerators are expensive assets whose usefulness and resale value can change as new chip generations emerge. Residual-value support may give lenders additional protection, although the precise scope of that protection depends on the transaction’s contractual terms.

Why banks are financing AI computing capacity

The deal reflects a shift in how financial institutions participate in the AI economy. Their role extends beyond providing working capital or underwriting conventional corporate debt. Banks and investment firms are increasingly being asked to structure financing around physical infrastructure, contracted demand and specialized equipment.

For AI developers, securing computing capacity is a strategic requirement. Training increasingly capable models and serving large numbers of users require access to advanced processors, high-bandwidth networking and data-center infrastructure. Building that capacity can require substantial upfront investment before the resulting services generate sufficient revenue.

An equipment-backed structure can help separate infrastructure funding from an AI developer’s direct ownership of every asset. However, leasing does not eliminate the underlying economic cost: the borrower or lessee still needs to meet its contractual obligations, while lenders remain exposed to the structure’s credit and collateral risks.

For financial institutions, the transaction creates exposure to the expansion of AI infrastructure through a defined financing arrangement. Blackstone’s participation also demonstrates how alternative asset managers can work alongside commercial and investment banks on capital-intensive technology projects.

Broadcom’s role in the AI infrastructure supply chain

Broadcom supplies the XPU accelerators and networking solutions involved in the transaction. Custom accelerators can be designed around particular workloads, while networking equipment connects processors and systems so that computing resources can operate together at scale.

This puts Broadcom in a different position from companies that primarily sell AI applications or rent general-purpose computing capacity. Its business serves the semiconductor and infrastructure layer, where demand depends on data-center investment, chip deployment schedules and customers’ long-term computing requirements.

The agreement also shows why financing and technology supply are becoming increasingly interconnected. Hardware availability alone does not guarantee deployment: customers and infrastructure partners must also secure capital, data-center capacity, electricity, cooling and networking.

A growing financial market, with execution risks

The scale of the transaction comes amid continued growth in AI-related data-center investment. The International Energy Agency projects that global data-center electricity consumption could rise from approximately 485 terawatt-hours in 2025 to 950 TWh by 2030, underscoring the physical resources needed to support expanding digital workloads. IEA

That demand outlook does not guarantee returns for individual infrastructure investments. Financing structures of this size depend on assumptions about future computing needs, equipment performance, contract durability and the ability of AI companies to sustain their spending.

The deal also raises questions about concentration risk. A substantial amount of capital is being directed toward one AI customer and a specific technology deployment. If demand changes, hardware becomes less competitive or deployment timelines slip, the parties could face financial and operational pressure.

The announcement does not disclose enough detail to assess the transaction’s ultimate cost to Anthropic or the returns available to its financing partners.

For the broader financial technology sector, the deal is less about digital payments or open banking than about the evolution of institutional capital markets. It demonstrates how structured finance can support a technology ecosystem whose growth depends on expensive physical assets, while leaving investors to evaluate credit exposure, collateral value and long-term demand.

The planned 2027 deliveries will be an important milestone, but the longer-term test will be whether the financed infrastructure generates enough sustained use to justify its cost.

Market Landscape

The $63 billion transaction highlights three developments in AI infrastructure finance:

• Structured finance for AI hardware: SPVs, equipment leasing and layered debt can help fund large computing deployments without requiring the end user to own every asset directly.
• Banks and alternative asset managers converge: Traditional financial institutions and private capital firms are collaborating on infrastructure financing that combines credit structuring with specialized investment expertise.
• Physical infrastructure remains a constraint: AI growth depends on processors, networking, power and data-center capacity—not just software innovation.

The IEA’s projection that data-center electricity demand could nearly double between 2025 and 2030 reinforces the scale of the infrastructure challenge. The investment opportunity, however, remains subject to execution, credit, technology-obsolescence and energy-supply risks.

Top Insights

• Bank of America, Citi and Morgan Stanley are arranging senior financing, while Blackstone provides committed second-lien financing for Anthropic’s AI infrastructure.
• The special purpose vehicle will purchase Broadcom XPU accelerators and networking equipment, then lease the hardware to Anthropic.
• Broadcom’s residual-value support is intended to address equipment collateral risks, although detailed contractual protections have not been publicly disclosed.
• Expected deliveries beginning in 2027 make execution timelines, hardware deployment and sustained demand important factors for the transaction’s success.
• The deal illustrates how structured finance and institutional capital are becoming important funding mechanisms for capital-intensive AI computing infrastructure.

Get in touch with our fintech expert