Broadcom has agreed to extend up to $42 billion to Anthropic for infrastructure and chip leasing, as reported by Reuters citing the developer's IPO prospectus.
This arrangement connects the semiconductor manufacturer with one of the leading AI firms through supply, leasing, and financing agreements.
Broadcom Takes on the Role of Lender for Anthropic
Anthropic outlined the terms of its partnership in a prospectus ahead of its potential IPO. Broadcom will provide a credit line of up to $42 billion to fund infrastructure expenditures.
Simultaneously, Broadcom will supply Anthropic with equipment and assist in developing a computing infrastructure utilizing its own solutions. By 2027, Anthropic is expected to become Broadcom's largest client in the specialized chip development sector.
In this scenario, Broadcom effectively plays multiple roles: equipment supplier, lessor, and lender. Reuters notes that this model is reminiscent of Nvidia's strategy, which also leverages financial resources to boost chip sales.
$42 Billion to Cover a Third of Major Contract
The credit facility is designed to cover roughly one-third of Anthropic's obligations under a five-year lease for tensor processing unit (TPU) computing power.
The total value of this contract is estimated at $125.2 billion. Starting in 2027, Anthropic will gain access to multiple gigawatts of next-generation TPU computing power developed by Google and Broadcom.
Funding will be structured through debt instruments that may potentially convert into Anthropic shares. However, the company has indicated it does not anticipate selling these securities until after its IPO.
This deal also serves Broadcom as a means to bolster future demand for its AI chips, with the company projecting around $115 billion in revenue from this segment in the fiscal year 2027 and $230 billion in 2028.
Anthropic Faces Significant Commitments
The agreement with Broadcom is just part of Anthropic's broader initiative to establish its own computing infrastructure.
According to another analysis of the company's prospectus, Anthropic aims to invest at least $518 billion over the next decade in AI infrastructure, with approximately 80% of these commitments being non-cancellable or requiring payment regardless of actual usage.
In addition to Broadcom, Anthropic has secured substantial agreements with Google, Amazon, and Microsoft. Its minimum commitments to Google total $111.1 billion, to Amazon $110 billion, and to Microsoft $31.4 billion.
Furthermore, the company has around $161.2 billion in leasing obligations with Broadcom, the majority of which are also non-cancellable.
Anthropic justifies these extensive expenses by emphasizing the need to ensure access to computing resources in advance, believing that a shortage of capacity, rather than demand for AI, could become the principal limitation on further development.
Risks Involved in the Arrangement
Anthropic acknowledges that having Broadcom serve as both equipment supplier and financial partner creates potential conflicts of interest.
For instance, Broadcom's decisions regarding pricing, equipment, and financing terms could impact Anthropic's ability to secure the necessary computing power. The company also warns that certain breaches of payment or other obligations could trigger significant portions of lease payments to come due early and simultaneously restrict access to the credit line.
This type of dependency is not exclusive to Broadcom; Google, Amazon, and Microsoft are also investors, suppliers of computing power, distributors, and competitors of Anthropic in the AI domain.
Additionally, Microsoft, Meta, Oracle, Amazon, and Alphabet have taken on approximately $1.09 trillion in future lease obligations, primarily for AI data centers, according to data released by Reuters in August.
Follow ForkLog on social media
Telegram (main channel) Facebook X