Mumbai: There was a time when an AI company needed engineers, servers and a reasonably good idea. Anthropic has apparently reached the stage where it needs something else too: a very large financial system attached to its computing system.
Anthropic’s latest IPO disclosures reveal that Broadcom has agreed to provide up to $42 billion in financing to help fund infrastructure spending tied to leased AI chips. The financing could cover roughly one-third of Anthropic’s $125.2 billion, five-year commitment for TPU computing capacity. The arrangement involves convertible notes, meaning the debt could potentially convert into Anthropic shares.
That makes this more than a conventional chip-supply agreement. Broadcom is simultaneously positioned as a hardware partner, equipment supplier and financier. In AI, apparently, the chip now comes with its own banking department.
Compute Has Become The Real Cost Of Intelligence
Anthropic’s expansion has been unusually aggressive.
In April, the company announced a multi-gigawatt agreement with Google and Broadcom for next-generation TPU capacity expected to begin coming online in 2027. Broadcom separately disclosed that Anthropic would access approximately 3.5 gigawatts through Broadcom as part of that broader TPU commitment.
Anthropic has also built a deliberately diversified computing strategy. Its models run across Google TPUs, Amazon Trainium and Nvidia GPUs, rather than depending entirely on one accelerator platform. The company says that diversification improves performance and resilience.
That strategy has a practical benefit: AI companies can shop around for the hardware best suited to different workloads.
The inconvenient part is that almost every shopping trip now costs billions.
The Numbers Have Become Almost Absurd
Anthropic’s financial disclosures show how quickly the economics have expanded.
The company reported approximately $4.6 billion in 2025 revenue, while its operating expenses reached about $12.65 billion. More than half of those operating expenses — around $7.33 billion — went toward compute and infrastructure, according to its IPO disclosures.
And the future commitments are considerably larger.
Anthropic has disclosed approximately $518 billion in future obligations connected to cloud, computing and infrastructure arrangements. That figure represents commitments rather than money already spent, which is an important distinction when discussing the scale of the business.
Meanwhile, Anthropic says its run-rate revenue had surpassed $30 billion by April 2026, up from roughly $9 billion at the end of 2025. It also said more than 1,000 business customers were spending over $1 million annually on an annualised basis.
The growth is real. So is the bill.
Broadcom Is Taking A Bigger Seat At The Table
For Broadcom, the arrangement potentially creates a powerful commercial loop.
Anthropic needs enormous amounts of computing capacity. Broadcom helps design and supply the underlying technology and is now also prepared to finance part of the infrastructure.
Anthropic is expected to become Broadcom’s largest compute customer in 2027, according to disclosures surrounding the arrangement. Broadcom has projected AI semiconductor revenue of approximately $115 billion in fiscal 2027 and $230 billion in fiscal 2028. Those are company projections, not guaranteed outcomes.
The positive case is fairly straightforward: financing can allow an AI company to secure computing capacity sooner, while semiconductor companies gain greater visibility into future demand.
That could accelerate model development, enterprise adoption and the construction of AI infrastructure.
But there is another side.
When Your Supplier Also Finances Your Hardware
Anthropic itself identifies a potential complication.
Its filing warns that Broadcom’s simultaneous role as a hardware supplier and financing partner creates potential conflicts of interest. The company also notes that decisions involving hardware pricing and supply could affect its ability to obtain enough computing infrastructure.
There are also conditions around the financing. Anthropic has placed cash into a restricted account for Broadcom’s benefit, and certain payment or performance defaults could make a substantial portion of its lease obligations immediately payable while restricting access to the financing facility.
None of this means the arrangement is inherently problematic. It does mean the relationship becomes considerably more intertwined.
And intertwined is one of those words that sounds elegant until someone has to read the footnotes.
AI Is Starting To Resemble Infrastructure Finance
The broader shift is difficult to miss.
Nvidia has already been working with financial institutions and infrastructure investors to help fund AI data centres and computing capacity. Broadcom’s Anthropic arrangement pushes that model further into the custom-chip ecosystem.
The emerging formula increasingly looks like:
AI model → enormous compute demand → chip supplier → long-term lease → financing → future AI revenue.
The advantage is speed. Companies can secure infrastructure without funding every requirement entirely from existing cash.
The risk is concentration.
If AI demand continues expanding rapidly, these arrangements could help build the infrastructure required to serve it. If demand, pricing or model economics change faster than the financing contracts do, the same commitments could become considerably less comfortable.
Anthropic’s story therefore isn’t simply about a $42 billion loan.
It is about something bigger: the cost of becoming an AI frontier company is increasingly being measured not only in models and engineers, but in years of committed computing capacity.
The machines may be intelligent.
The contracts, unfortunately, remain very human.
Read More: Amazon Eyes $8 Billion Nvidia Chip Sale
