Bank of America, Blackstone, Citigroup and Morgan Stanley will provide $63 billion in financing to fund the purchase of artificial intelligence chips and networking equipment from Broadcom for Anthropic, underscoring the growing role of debt markets and alternative asset managers in financing the AI infrastructure boom.
The transaction ranks among the largest financing arrangements linked to the expansion of AI computing capacity and highlights how the industry’s enormous infrastructure requirements are creating new opportunities for lenders while exposing them to the risks associated with rapidly evolving technology and asset valuations.
The financing will be channeled through a newly formed special purpose vehicle that will purchase Broadcom’s AI accelerators and networking equipment and lease them to Anthropic, the developer of the Claude family of AI models. Deliveries are expected to begin in 2027 under a previously announced strategic compute capacity agreement involving the two companies.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Bank of America, Citigroup and Morgan Stanley are arranging the senior, or first-lien, financing, while Blackstone will provide the committed portion of the second-lien financing. First-lien lenders have priority over second-lien creditors in claims against the relevant collateral, making the structure an important determinant of how potential losses would be distributed if the transaction encountered financial difficulties.
Two sources familiar with the matter told Reuters the first-lien loans or senior debt would total approximately $43.5 billion, representing about 69% of the overall financing. The companies did not disclose the complete breakdown of the loans in their announcement.
Broadcom is providing what the companies described as “residual value support” to lenders, effectively offering protection against a decline in the value of the financed equipment. Such arrangements can improve the financing terms available to borrowers by reducing lenders’ exposure to losses if the underlying assets are worth less than anticipated.
The structure indicates that AI infrastructure financing is evolving beyond conventional corporate borrowing. Rather than relying entirely on Anthropic’s balance sheet or raising all the required capital directly, the arrangement uses a separate financing vehicle to acquire equipment and lease it to the AI company.
This approach allows the parties to connect the financing more directly to identifiable assets and their expected economic use. The special purpose vehicle purchases the equipment, lenders provide the capital, and Anthropic obtains access to computing infrastructure through leasing arrangements.
However, the value of the collateral is key to the risk assessment. AI accelerators and networking equipment are expensive assets, and their economic value is determined by their useful life, demand from other potential buyers, technological advances, and the ability to generate sufficient revenue from the computing capacity they support.
Therefore, Broadcom’s residual value support is a significant feature of the transaction. By helping protect lenders against potential declines in equipment value, the chipmaker assumes some exposure that might otherwise fall more heavily on creditors. The precise extent of that exposure depends on the terms of the support arrangement, which were not disclosed.
The involvement of multiple financing layers also matters. Senior lenders benefit from first priority over the relevant collateral, while second-lien creditors rank behind them. Blackstone’s commitment to the second-lien portion places it in a structurally more junior position than the first-lien lenders.
The arrangement offers banks exposure to the financing needs of a rapidly expanding technology sector without necessarily requiring them to fund the entire transaction through a conventional corporate loan to Anthropic. On the other hand, it represents for Blackstone participation in the financing of AI infrastructure through a structure that includes a more subordinated layer of debt.
Yet the presence of collateral and residual value support does not eliminate risk. If the equipment loses value faster than expected, computing demand weakens, or the lessee encounters financial difficulties, lenders could still face losses depending on the strength and scope of the contractual protections.
Moody’s has described Broadcom as one of the strongest large technology companies but warned that risks could increase if multiple transactions are undertaken over a short period to meet the substantial funding needs of frontier AI laboratories.
That concern points to a broader issue in the AI investment cycle: the capacity of companies and their financing partners to support increasingly large infrastructure commitments without concentrating too much risk in a small number of borrowers, technology suppliers and assumptions about future demand.
Frontier AI Expansion Tests The Financing Model
The $63 billion arrangement follows an earlier agreement under which Apollo and Blackstone agreed to finance a $35 billion expansion of Anthropic’s AI computing capacity involving Broadcom. Broadcom disclosed in a filing that its backstop in that transaction could reach $29 billion.
Taken together, the arrangements underpin the scale of capital required to expand computing capacity for frontier AI systems. The investment extends beyond purchasing processors to include the networking equipment and supporting infrastructure needed to operate large computing clusters.
Technology companies are racing to secure capacity for complex AI workloads, including systems that can perform multistep tasks and handle more demanding computational processes. Access to sufficient computing infrastructure has consequently become a major determinant of how quickly AI developers can train and deploy their models.
The financing model also distributes exposure across a broader group of participants. AI developers secure computing capacity, chipmakers sell equipment and may provide contractual support, while banks and alternative asset managers supply capital. This arrangement is touted to accelerate infrastructure deployment by allowing investment to proceed without requiring a single company to finance every component directly.
But it also creates financial links between parties whose fortunes are increasingly tied to the same underlying assumption: that demand for advanced AI computing will remain strong enough to justify the capital committed today.
If demand grows as expected, long-term leasing arrangements can help AI developers secure infrastructure while providing lenders with contractual repayment streams and equipment as collateral. If demand falls short, however, the industry could face pressure from underutilized computing capacity, weaker equipment resale values and financing commitments that remain in place even as the economics of AI services deteriorate.
Anthropic’s own financial position will be relevant to that assessment. The company filed for a potentially record-breaking initial public offering on Monday, even as disclosures of steep losses and security concerns threatened to test investor appetite for the AI heavyweight.
An IPO could give public-market investors greater visibility into Anthropic’s financial performance, capital requirements and ability to turn growing demand for its models into sustainable returns. It would also place additional scrutiny on the relationship between infrastructure spending, revenue growth and the cost of maintaining a competitive position in frontier AI.
The $63 billion financing is not interpreted to mean that Anthropic will struggle to meet its obligations or that the AI infrastructure market is overbuilt. Nor does Broadcom’s support mean lenders are insulated from every possible loss.