Anthropic’s IPO: A Great Business at Any Price? Absolutely Not.

Would I want to own a piece of Anthropic? Yes. Would I buy its IPO without looking hard at the price? Absolutely not. Those are different questions, and the gap between them is where investors could get hurt. As of November 3, 2026, Anthropic has filed to list its Class A shares on Nasdaq under ANT. It updated its prospectus yesterday with third-quarter results. But there is still no announced offer price, share count, or listing date. Anyone telling you this is already a bargain is pricing a stock that does not yet have a price.

First, the upside—and it is substantial. Claude is not just a chatbot people enjoy trying. Anthropic sells model access to businesses and developers, much of it metered by usage. Revenue reached nearly $17 billion in the July-to-September quarter, up from about $1.4 billion a year earlier. More than 11,900 customers generated at least $100,000 each over the preceding twelve months, versus more than 950 a year before. That is the strongest part of this pitch: customers are putting serious money behind the product. If Claude becomes embedded in how companies write software and run workflows, today’s revenue could be the start of something enormous.

Now for the number that will dominate the headlines: Anthropic lost $104 billion in the first nine months of 2026. That sounds like a company setting money on fire with a flamethrower. It is also a terrible shorthand for its operating economics. Roughly $94 billion of the loss was a non-cash accounting charge tied to the rising value of convertible financing. The business itself posted an operating loss of about $1.3 billion over those nine months, while operations generated about $6.6 billion in cash. But do not swing to the opposite fantasy and call it reliably profitable: a $974 million operating profit in the second quarter became a $567 million operating loss in the third.

The reason I hesitate is compute. The updated filing puts Anthropic’s signed contracts at about $895 billion, substantially all for computing and infrastructure. Read that carefully: it is not a bill due tomorrow, and some agreements are cancellable or conditional. Still, about $69.8 billion falls due in the twelve months after September 30—roughly four quarters of sales at the latest quarter’s pace. Anthropic says expensive short-term computing helped push third-quarter operating results back into the red. My concern is not that demand is imaginary. It is that demand might grow spectacularly while the cost of serving it grows almost as spectacularly. Revenue is not a moat if every new customer needs another costly block of capacity.

There is a second, awkward dependency. Amazon, Google and Microsoft provide routes through which customers buy Claude; their cloud marketplaces accounted for about 44% of revenue in the first nine months. Major technology partners also supply computing and invest in Anthropic. Those relationships help it scale, but they make the company dependent on firms with substantial leverage over its distribution or infrastructure. And AI pricing is not sacred: the filing says new-model price cuts and contract discounts reduced revenue by about $600 million between the second and third quarters. My bearish scenario is not ‘nobody wants Claude.’ It is ‘everybody wants Claude, but competition and infrastructure suppliers capture too much of the value.’

Then there is governance. Public buyers get Class A shares, but the Long-Term Benefit Trust elects most directors, while a founders’ vehicle holds a special share carrying 50.1% of the vote on other matters. In practice, I would treat these as shares with very little influence over who runs the company. Anthropic’s public-benefit structure may be principled; I respect that. But if management makes a costly strategic bet, ordinary shareholders have limited power to change course. There are other risks, too: copyright litigation, AI safety failures, regulatory friction, and a dispute over a federal supply-chain-risk designation. Those are not reasons to declare disaster inevitable. They are reasons to demand a price that leaves room for things to go wrong.

So what price would make sense? We cannot answer precisely until Anthropic publishes the offer terms. Its May private funding round valued it at $965 billion, according to the company. That is not the IPO valuation. For perspective, that private figure is about fourteen times the latest quarter’s revenue annualized—a calculation, not a forecast. At anything like that valuation, I would want convincing evidence that growing usage produces durable profits after compute costs, stock compensation and future dilution. I would watch margins, capacity spending, and whether customers keep paying as prices fall. My verdict: Anthropic may be one of the defining businesses of this era. That makes it worth studying intensely. It does not make an unpriced IPO an automatic buy. Admire the product; interrogate the economics; and do not let the excitement write the cheque.