Anthropic’s confidential IPO prospectus, obtained by Reuters, reveals that Broadcom has agreed to lend the AI company up to $42 billion to help fund its infrastructure buildout. The arrangement makes Broadcom the third major technology company, following Amazon and Microsoft, to both finance and supply computing capacity to Anthropic simultaneously, a structure that has drawn mounting scrutiny across the AI industry over the past several months.
The concern is straightforward. Each of these companies provides capital to Anthropic, which Anthropic then uses to purchase computing services from that same company. Anthropic’s own spending with Broadcom is expected to grow so large that the AI lab will become Broadcom’s single largest compute customer by 2027. The relationship extends beyond simple chip purchases as well. Broadcom is also providing Anthropic with equipment leasing and financing arrangements, according to Reuters, deepening the financial interdependence between the two companies beyond a standard customer relationship.
This builds directly on a partnership announced in April, when Anthropic revealed a three-way arrangement with Broadcom and Google under which Google will supply its own Tensor Processing Unit capacity to Anthropic beginning in 2027. Broadcom, notably, is the company that designs Google’s TPUs, meaning Broadcom now sits on multiple sides of Anthropic’s compute supply chain simultaneously, as both a direct lender and as the designer of chips Anthropic will access through a separate cloud partner.
Broadcom is not acting in isolation here. Nvidia and AMD have each extended similar funding arrangements to their own major AI lab customers, including both OpenAI and Anthropic, with those labs then using the capital to pay for access to the same companies’ high-performance chips. The pattern has become common enough across the industry that it now has a name, circular financing, and it has become one of the more closely watched risk factors in the entire AI infrastructure buildout.
The systemic concern is what happens if any single link in that chain breaks. If one major AI lab or chip supplier in these circular arrangements were to stumble, whether through a funding shortfall, a demand slowdown, or a failure to meet compute obligations, the interconnected nature of these deals raises the risk of a chain reaction spreading across multiple companies at once, with implications extending well beyond the AI sector into the broader equity markets that have leaned heavily on AI-driven earnings growth over the past several years.
For investors tracking this space, the Broadcom-Anthropic arrangement is best understood as the latest data point in a theme we’ve tracked closely across several recent stories. Nvidia’s record $150 billion buyback expansion drew scrutiny partly because of circular financing concerns raised on its own earnings call. BlackRock’s $12.3 billion bond offering to fund Meta’s data center campus and CoreWeave’s aggressive capital expenditure guidance both reflect the same underlying dynamic, debt and vendor financing being used at massive scale to fund an AI buildout whose ultimate revenue payoff remains, in some cases, still unproven.
For investors in the small and microcap space, the direct read-through is more nuanced than it might first appear. Smaller companies supplying components, materials, and specialized hardware into this ecosystem benefit from the sheer scale of spending these arrangements represent, $42 billion from a single lender to a single customer is a genuinely enormous demand signal. But that same scale is precisely what makes the systemic risk real. A supply chain this interconnected and this reliant on vendor financing rather than organic cash flow is more fragile than the headline spending figures alone suggest, and investors evaluating smaller AI infrastructure suppliers should weigh that structural risk alongside the genuine demand opportunity it represents.
