Iran War: Why Are They BEGGING Trump for Protection?

The number that will define Anthropic’s IPO roadshow is not the revenue figure, impressive as $17.0 billion in a single quarter is. It is $895.4 billion, the total disclosed in the amended S-1 filed November 2, covering purchase commitments and leases that are, in the company’s own words, “substantially all” for compute and infrastructure.

The Bull Case

Start with what the commitment buys. Much of the spending relates to Anthropic Direct, the company’s own dedicated infrastructure, which it says will let it replace expensive short-term compute with cheaper long-term capacity as it comes online. That argument has teeth. The Q3 operating loss was small against $16.956 billion of revenue, about 3.3%, and Anthropic said it was “primarily driven by $5.0 billion of compute under short-term commitments.” The loss is, in other words, a transition cost, not a structural one.

The revenue trajectory is still extraordinary by any honest measure. In Q3 alone, the company took in $17.0 billion, against $1.4 billion a year earlier. Year-over-year growth exceeded 1,000%. Adjusted operating profit was $2.5 billion for the quarter, suggesting the underlying business, stripped of transition-period compute costs, is already generating real cash. Long-term contracts with Amazon, Google, and Microsoft also lock in supply that competitors cannot easily access, which matters when GPU availability is the binding constraint on AI product development.

Named minimum commitments include at least $121.3 billion to Google, at least $110.0 billion to Amazon, and at least $44.9 billion to Microsoft. Those are not merely vendor relationships. They are strategic alignments with the three largest cloud platforms, each of which has a direct incentive to see Anthropic succeed on their infrastructure.

The Bear Case

The bullish read requires one assumption to hold: that revenue keeps compounding fast enough to service a commitment schedule that front-loads heavily. Anthropic expects about $447.5 billion to be incurred before December 31, 2030, with approximately $69.8 billion falling due in the 12 months after September 30. Against that near-term obligation, the full bill is more than six times the roughly $131.9 billion of cash, cash equivalents, and short-term investments Anthropic held at the end of September.

And the growth deceleration in the amended filing is impossible to ignore. Revenue rose about 47% sequentially in the three months ended September 30, up from $11.555 billion in the prior quarter. One quarter of deceleration does not make a trend, but the direction is wrong at exactly the moment commitments are accelerating. Compute and infrastructure costs climbed about 77% over the same period, rising from $5.558 billion in Q2 to $9.863 billion in Q3 while revenue rose 47%. Cost growth outpacing revenue growth is the arithmetic that ends companies.

The cost pressure is not Anthropic’s alone to absorb. Broadcom’s custom silicon division sits at the center of the hyperscaler capex surge that is driving compute costs higher across every AI lab.

The filing itself acknowledges the risk without softening it. “Our ability to meet the substantial and growing financial obligations of these requirements to secure compute depends on our ability to generate significant revenue growth,” the prospectus states. That sentence is the entire bear case in one line. Of the $895.4 billion in contractual obligations, approximately $213.8 billion relates to arrangements that are subject to termination rights or contingent conditions prior to commencement. The remaining obligations are largely fixed regardless of what revenue does. The company also disclosed that revenue growth between Q2 and Q3 was partly offset by approximately $600 million from price reductions introduced with new model versions and contractual discounts, suggesting competitive pricing pressure is already compressing realized revenue.

Where the Evidence Leads

The bull and bear cases share the same hinge point: whether 47% sequential growth is a temporary trough or the new ceiling. If Anthropic can re-accelerate toward its prior rate, the long-term infrastructure position looks like a durable cost advantage. If 47% is where growth stabilizes, the company has locked itself into an obligation schedule that revenue cannot support without perpetual external financing.

The adjusted profitability number, $2.5 billion in operating profit before the short-term compute charges, is the most important figure in the filing for long-term bulls. It shows the model works when infrastructure costs normalize. The bear responds that normalization depends on the long-term compute coming online as promised, at the projected cost savings, on schedule, across contracts with Broadcom, SpaceX’s AI unit, Google, Amazon, and Microsoft simultaneously.

The normalization thesis depends entirely on whether Anthropic can escape the cloud markup. Anthropic’s push to lease capacity directly from independent data center operators is the structural mechanism behind the projected cost savings.

What Could Change the Debate

Anthropic has applied to list on Nasdaq under the ticker ANT, but the filing does not specify an IPO date. Pricing will reveal whether institutional investors accept Anthropic’s own framing of the compute commitment as a strategic asset rather than a liability. Any Q4 revenue signal suggesting re-acceleration above 47% sequential growth would decisively strengthen the bull case. Deterioration below that rate, particularly if accompanied by further price reductions to hold volume, would validate the bear’s concern that demand is more price-sensitive than the revenue scale implies.

The Verdict

The evidence today favors cautious skepticism, not rejection. Anthropic’s adjusted economics are real, its counterparties are credible, and its year-over-year growth remains historic. But the amended filing’s combination of decelerating sequential growth, rising cost intensity, and a near-term payment schedule that dwarfs current liquidity means prospective investors are buying a forecast as much as a business. The asking price will need to reflect that asymmetry.