Broadcom to provide Anthropic with $42 billion loan to cover expenses and fund next-generation chips
Read more
Ventureburn
ventureburn.com

Broadcom to provide Anthropic with $42 billion loan to cover expenses and fund next-generation chips

A recent leak of Anthropic's initial public offering (IPO) prospectus demonstrated how expensive it is to build super-powerful artificial intelligence. Despite impressive revenues, the company faces cash flow gaps while striving to secure sufficient computing power to maintain a competitive edge.

Due to financial difficulties, Anthropic was in urgent need of capital to support large-scale hardware. To address this colossal infrastructure challenge, an unprecedented move was made: Broadcom agreed to provide Anthropic with a substantial loan of $42 billion as part of a major financial deal.

This deal between Broadcom and Anthropic goes beyond simply providing funds; it is highly strategic and was included in Anthropic's recent IPO prospectus. Broadcom is effectively financing the startup's infrastructure costs through a large convertible debt transaction.

Anthropic is legally obligated to enter into a five-year lease agreement for Tensor Processing Unit (TPU) capacity worth an incredible $125.2 billion. This volume of equipment is enormous, and the large loan from Broadcom covers approximately one-third of these obligations.

Together with Google, Broadcom will supply these next-generation TPUs starting in 2027. This will allow Anthropic to gain the necessary computing power to compete with OpenAI, while Broadcom secures a stable and significant revenue stream.

This approach is similar to the tactic previously employed by NVIDIA. In recent years, NVIDIA has used its strong balance sheet to finance ambitious AI startups. These startups received large capital injections, which they then spent on purchasing NVIDIA GPUs, creating a closed loop of expenditure.

Broadcom is implementing the same strategy, using its financial might to artificially stimulate demand for its own chips. By providing this large line of credit, Broadcom ensures that Anthropic will be long-term tied to its hardware. By 2027, Anthropic is projected to become Broadcom's largest consumer of computing power.

Anthropic is preparing for a public listing that could value the company at an astronomical $2 trillion. However, there is increased caution on Wall Street regarding such interconnected technology deals. Anthropic acknowledged in its documentation that Broadcom's dual role—as both a critical hardware supplier and the primary lender—creates obvious conflicts of interest.

If Anthropic fails to meet its targets, the situation could become extremely difficult, as a significant portion of the lease obligations may suddenly come due. The race in AI infrastructure is narrowing down to a few large institutional players requiring gigawatts of energy and thousands of advanced chips to create frontier AI models.

Broadcom's willingness to allocate $42 billion demonstrates how profitable the custom AI chip market has become and shows how startups are attempting to circumvent the dominance of large industry players.

Similar stories

Anthropic details IPO valuation and billion-dollar costs for artificial intelligence in its prospectus
Read more
olhardigital.com.br

Anthropic details IPO valuation and billion-dollar costs for artificial intelligence in its prospectus

Anthropic expresses the belief that artificial intelligence (AI) has the potential to promote a global economic transformation even deeper than industrialization, electricity, and the internet. However, the prospectus for its Initial Public Offering (IPO) reveals the considerable financial challenge required to sustain this vision.

According to the document, the company recorded a net loss of US$ 42 billion (equivalent to about R$ 227 billion) in 2025 and projects future commitments of US$ 518 billion (approximately R$ 2.8 trillion) dedicated to cloud, computing, and infrastructure next year. These figures are part of the IPO prospectus, as reported by Reuters.

The public offering could value Anthropic at over US$ 2 trillion (about R$ 10.8 trillion), more than double the valuation the company itself estimated in May, when the value was US$ 965 billion (approximately R$ 5.2 trillion). This move could serve as an important indicator for the market regarding the amount investors are willing to pay for leading AI companies.

It is important to note that not all net loss corresponds to expenses actually incurred by the company. Of the nearly US$ 42 billion loss, about US$ 34 billion (approximately R$ 183.6 billion) is related to an accounting expense resulting from the projected increase in the value of financing that can be converted into Anthropic shares. Thus, a substantial portion of the loss recorded in the result is tied to the accounting of these financial instruments, and not directly to operational costs.

Despite this, infrastructure spending is vast, and the company anticipates even larger commitments to support its expansion.

More information about the prospectus

The prospectus also highlights risks associated with revenue concentration. In 2025, almost a quarter of Anthropic's revenue came from just two clients. Furthermore, the company warns that many of its largest clients do not have long-term agreements and may choose to decrease or suspend their spending.

This aspect is among the risks investors must consider if the company proceeds with its public offering. The IPO is expected to occur after the midterm elections in the United States in November, according to previous Reuters reports. The offering would place public investors directly into an AI competition that, until now, was mainly funded by venture capital funds, sovereign wealth funds, and large technology corporations.

Anthropic competes directly with OpenAI. Both companies compete for corporate clients, specialized professionals, and influence in Washington (USA). OpenAI submitted its IPO application confidentially in June and may enter the market in early 2027. Anthropic also faces competition from xAI, Google, and Meta in developing AI infrastructure.

Analysts predict that the first major AI company to go public will help establish valuation standards for the entire sector and attract investors looking for a direct way to bet on technological growth.

The prospectus also reveals an internal contradiction at Anthropic. While the company promotes accelerated AI expansion, internal research has pointed to evidence that progressively more autonomous models can exhibit unpredictable and potentially harmful behaviors in controlled tests. The studies mentioned in the document identified actions such as code sabotage, assistance in fraud, and data manipulation, raising concerns about how to keep increasingly powerful systems under control during the acceleration of commercial implementation.

Dario Amodei, CEO of Anthropic, has already argued that the global AI community should moderate the pace of new feature releases to address these concerns. Nevertheless, Anthropic recently launched its new Opus 5.5 model in response to OpenAI's advancement after the launch of GPT-6 Astra, while Anthropic itself prepares for its possible IPO.

The projected valuation of over US$ 2 trillion would position Anthropic among the most valuable technology companies on the planet. This value exceeds the US$ 965 billion estimate made by the company itself in May by more than double, demonstrating the magnitude of expectations placed on AI growth. The IPO will occur during a period of intense scrutiny over the vast resources that technology companies are allocating to building AI infrastructure. For Anthropic, the documents presented to investors clearly illustrate the scale of the investment: billions in revenue are already accompanied by tens of billions in expenses and hundreds of billions in future commitments for computing and infrastructure.

Popular