Anthropic's $2 trillion IPO bet: an $8 billion operating loss matters more than the $42 billion headline

Anthropic's IPO prospectus reveals a company that is growing fast and losing a lot of money. According to an analysis by Stefon Walters of The Motley Fool, published on Yahoo Finance, the maker of Claude grew revenue by 1,050% in 2025, to $4.6 billion. Its net loss for the year was $42 billion. The same analysis says the market could value Anthropic at up to $2 trillion when it lists. That would edge out SpaceX as the largest IPO in history.

The headline figure needs context. Per the analysis, about $34 billion of the net loss came from an accounting charge. The operating loss, which counts only the cost of running the business, was about $8 billion. That is still large. Anthropic reported $12.65 billion in total operating expenses against $4.6 billion in revenue. In other words, it spent roughly $2.75 in operating costs for every dollar it brought in. Compute was the biggest item: $7.3 billion in 2025, or 58% of operating expenses. That compares with $2.5 billion in 2024 and $400 million in 2023, an eighteenfold rise in two years.

The valuation makes the gap more striking. At $2 trillion, Anthropic would trade at more than 400 times its 2025 revenue. Mature software companies are usually valued at a small fraction of that multiple. The Motley Fool puts it plainly: investors accept "all burn, no profit" from young, high-growth companies, but that patience has limits when the price tag is $2 trillion. Until now, Anthropic's price has been set in private funding rounds. A listing would be the first time public shareholders test it against audited, public financial statements every quarter.

The risk section is unusual too. Most prospectuses warn about competition, regulation or reliance on suppliers. According to the analysis, Anthropic's lists the difficulty of keeping autonomous AI models contained. It also acknowledges that advanced AI could pose "catastrophic or existential risks to humanity", including the possibility that AI systems put their own interests above human ones. Including this is partly legal caution. But it also turns a philosophical debate into a formal risk disclosure that securities regulators and shareholders can hold the company to.

Our reading: this IPO is the first real test of the economics of the AI boom, not just of Anthropic. The number to watch is not revenue growth, because 1,050% will not repeat. It is the ratio of compute spending to revenue. If compute keeps rising as fast as sales, the model never becomes profitable. If inference costs fall and enterprise contracts keep growing, operating leverage will appear. Inference is the cost of running a model for customers, as opposed to training it. In the short term, the risk is real. A market that prices in perfection can swing sharply, and a correction could hit funding for the whole sector. For the long term, our view is more constructive. A frontier lab that has to publish its accounts and its safety risks is better for everyone than one that only answers to private investors. If the compute being bought today produces systems that speed up science and medicine, the losses will look like an investment. Fable 5 tops our quality rankings, so the technology is there. The prospectus has to show the business can be sustained.