Anthropic IPO: $518B Compute Bill Tests a $2 Trillion Pitch

A Blockbuster Listing With a Heavy Anchor

Anthropic is reportedly preparing to go public with a valuation pitch of around $2 trillion. Yet the most consequential number in its leaked prospectus may be a different one: $518 billion. That is the size of the long-term AI infrastructure commitments the company has made, and it sits uneasily beside a business that is still deeply unprofitable.

According to the leaked filing, Anthropic posted a $42 billion net loss on just $4.6 billion of revenue in 2025 2. The listing is reportedly targeted for mid-November, with an investor day scheduled for October 14 where those figures are expected to draw pointed questions 2.

The Asymmetry at the Heart of the Deal

The $518 billion is spread across Anthropic's three big cloud partners: Google, Amazon and Microsoft 2. Roughly 80% of those commitments are reportedly non-cancelable 2. In practical terms, most of that compute bill is locked in regardless of how demand develops.

The revenue side is far less secure. The prospectus reportedly concedes that enterprise customers can walk away 1. That creates a structural mismatch. Anthropic's costs are fixed and contractual. Much of its income depends on corporate buyers who keep the option to switch models, renegotiate, or cut back spending 1.

This asymmetry is the real story beneath the headline valuation. A company with flexible costs and sticky revenue can survive a downturn. A company with sticky costs and flexible revenue carries leverage, even if that leverage does not look like conventional debt on a balance sheet. Non-cancelable compute contracts behave much like borrowing: they must be paid whether or not the customers show up.

Reading the Numbers

Set against each other, the figures are stark. Annual revenue of $4.6 billion is less than 1% of the committed infrastructure spend. The $42 billion loss is more than nine times revenue 2. For a $2 trillion valuation to make sense, investors would need to believe that revenue can scale by orders of magnitude, and fast enough to stay ahead of obligations that have already been signed.

That is not an impossible bet. Frontier AI labs have argued that compute capacity is the binding constraint on growth, and that locking in supply early is a competitive necessity. On that view, the commitments reflect conviction rather than recklessness. Still, the burden of proof falls on revenue growth and customer retention. Those are exactly the areas where the prospectus reportedly acknowledges weakness 1.

The concentration of counterparties also matters. Google, Amazon and Microsoft are infrastructure suppliers, and Google and Amazon have also been major investors in Anthropic. Public shareholders will want clarity on how those relationships shape pricing, terms and Anthropic's room to maneuver if demand softens.

Diverging Paths Among AI Leaders

Anthropic's sprint toward public markets contrasts with the reported strategy at OpenAI. OpenAI is said to be pulling back on its own IPO plans. It is instead raising at least $30 billion more in private capital at a $1.4 trillion valuation 2.

The split suggests the two leading labs are reading the same market differently 2. Going public gives Anthropic access to a deep pool of capital and a liquid currency for future deals. It also brings quarterly scrutiny, mandatory disclosure, and exposure to sentiment swings. Staying private lets OpenAI avoid that spotlight, at the cost of relying on a narrower set of very large backers. Both choices are unfolding amid rising scrutiny of AI safety 2, which adds regulatory and reputational uncertainty to any long-horizon forecast.

What to Watch on October 14

The investor day is likely to focus on a few pressure points:

• Revenue durability: how much enterprise revenue is under multi-year contract, versus usage that can disappear quickly.
• Commitment schedule: how the $518 billion is phased over time, and what the roughly 20% of cancelable obligations actually covers.
• Path to margin: whether inference costs are falling fast enough to turn growth into gross profit.
• Partner dynamics: how intertwined the investor and supplier relationships with the cloud giants really are.

The Verdict

Anthropic's prospectus, as reported, describes a company that has bet heavily on its own future demand. The $2 trillion pitch rests on the idea that AI adoption will grow into the infrastructure already reserved. That may prove right. But the gap between fixed obligations and revenue that customers can withdraw is a genuine risk, and it is not a minor footnote 12.

Investors considering the offering should treat the $518 billion less as a sign of scale and more as a commitment that must be paid. The valuation question is not whether Anthropic can build impressive models. It is whether its customers will stay long enough to pay for the compute it has already promised to buy.