Anthropic's IPO filing reveals it owes up to $161 billion for chips it hasn't fully used yet

Anthropic's IPO prospectus disclosed a $125.2 billion, five-year commitment to lease AI chip computing capacity, alongside an $8 billion operating loss in 2025 and a $42 billion financing arrangement with Broadcom to help fund the chip leases.

Why it matters: IPO filings require far more financial transparency than private companies typically disclose. This gives the public its clearest look yet at exactly how much it actually costs to run a company at the frontier of AI development.

Broadcom has agreed to lend Anthropic up to $42 billion through convertible notes, financial instruments that could eventually convert into Anthropic equity. That financing could cover roughly a third of Anthropic's massive chip-leasing commitment.

Why it matters: a loan structured to potentially convert into equity ties Broadcom's own financial fortunes directly to Anthropic's long-term success, not just a one-time lending transaction. Broadcom is also projecting its own AI semiconductor revenue to roughly double between fiscal 2027 and 2028, making this relationship central to both companies' futures.

The $125.2 billion commitment covers a five-year lease of TPU, or Tensor Processing Unit, computing capacity. TPUs are Google's custom AI accelerator chips, broadly comparable in role to Nvidia's GPUs but built specifically for AI workloads, with Google and Broadcom jointly supplying the chips Anthropic is leasing.

In real life it's the AI-era equivalent of leasing an entire fleet of specialized factory equipment years before you're sure how much of it you'll actually need running at full capacity.

Why it matters: leasing rather than owning the chips reflects how fast AI hardware improves. Locking in access to computing capacity, rather than owning hardware that could be outdated within a few years, has become the industry's default approach to scaling.

By the numbers: Anthropic reported an $8 billion operating loss in 2025, more than five times the prior year's loss. Computing infrastructure costs, the chips and data centers needed to run its AI models, accounted for more than 91% of that loss.

Why it matters: that concentration shows the loss isn't coming from scattered inefficiency across the business. It's overwhelmingly the direct cost of the computing power required to train and serve AI models at the scale Anthropic is now operating.

Yes, dramatically. Revenue grew from $400 million in 2024 to $4.6 billion in 2025, roughly a 1,050% increase. Momentum accelerated further into 2026, with second-quarter revenue alone reaching $11.5 billion.

Why it matters: that pace of growth is genuinely rare for a company of any size. It's part of why investors are reportedly willing to support a valuation above $2 trillion, even alongside billions in losses.

By the numbers: the company has committed to spending $518 billion on computing infrastructure over the next decade, and roughly 80% of those commitments are non-cancelable or require payment regardless of whether the capacity actually gets used.

Why it matters: that non-cancelable structure is a significant financial risk. If AI demand ever slows or Anthropic's own growth stalls, the company is still contractually obligated to keep paying for infrastructure it committed to years in advance.

Reporting indicates Anthropic is seeking a valuation above $2 trillion, with an offering that could raise around $100 billion, expected as soon as November 2026. Nvidia, which previously invested $10 billion in the company, could reportedly serve as an anchor investor.

Why it matters: a $2 trillion target valuation would place Anthropic among the most valuable companies in the world at the moment of its public debut, an extraordinary milestone for a company that didn't exist before 2021.

Yes. The prospectus reportedly devotes more than 80 pages to risk factors, including warnings that advanced AI models could pose what the filing describes as catastrophic or existential risks to humanity.

Why it matters: that's an unusually candid disclosure for a company to include in its own pitch to investors. It reflects how seriously Anthropic, a company built around AI safety research from its founding, treats those risks even as it asks the public markets to bet on its growth.

Background: Anthropic was founded in January 2021 by siblings Daniela and Dario Amodei, along with several colleagues who left OpenAI over concerns about the company's commitment to safety. Dario Amodei had previously led research at OpenAI and co-invented reinforcement learning from human feedback, a technique now foundational across the AI industry. Anthropic was structured from the start as a public benefit corporation built around AI safety research.

Why it matters: that origin story helps explain the unusual 80-plus pages of existential-risk disclosures in the IPO filing. A company founded specifically because its own leaders worried about AI safety at a rival lab is, almost by design, going to take those risk disclosures more seriously than a typical tech IPO would.

By the numbers: Anthropic was valued at roughly $965 billion in a Series H funding round in May 2026, just months before reportedly targeting a valuation above $2 trillion for this IPO.

Why it matters: more than doubling in valuation within the same calendar year, while simultaneously posting billions in losses, says less about near-term profitability and more about how much investor confidence currently rides on AI infrastructure spending continuing to pay off over a much longer time horizon.

• This filing is one of the clearest public windows yet into what it actually costs to compete at the frontier of AI, information that was previously mostly private.
• The scale of non-cancelable infrastructure commitments is worth watching across the whole AI industry, not just Anthropic, since similar lease structures are becoming standard among major AI labs.
• Rapid revenue growth and a rapidly growing operating loss are happening together here, a pattern worth understanding if you're trying to make sense of AI industry economics broadly.
• If Anthropic's IPO proceeds as expected around November 2026, it will become one of the most closely watched public offerings of the year, given both its scale and its direct ties to the broader AI infrastructure boom.