Anthropic’s planned stock market debut constitutes “an unprecedented test of investor credulity,” independent research firm New Constructs said in a report titled “Anthropic is the Most Ridiculous IPO of 2026.” The company estimates the AI company’s equity at just $144 billion. That’s 93% below the $2 trillion valuation that Anthropic is reportedly targeting.
The report was released on October 6. Since then, Anthropic has told investors that its third-quarter revenue reached $16.8 billion. Bloomberg reported that the listing could take place as soon as mid-November. The debate over the valuation of Anthropic’s IPOs is entering its decisive weeks.
What it would take for $2 trillion
New Constructs sells stock market research to investors. He used a reverse discounted cash flow model, which works backwards from a price to the profits needed to justify it.
To be worth $2 trillion, the company claims, Anthropic would immediately need a 10% after-tax operating margin and maintain it for a decade. Revenue is also expected to grow 1,000% this year, 100% in 2027, and 56% annually from 2028 to 2035. The report puts Anthropic’s operating margin for 2025 at minus 175%.
In this case, Anthropic’s after-tax operating profit would reach $358 billion in 2035. That’s “2.0 times NVIDIA’s $184 billion and 2.6 times Apple’s $136 billion” in current profit, the report said. Revenues would reach $3.6 billion. This represents approximately 1.4 times the combined revenue of Microsoft, Apple, Alphabet, Amazon, Nvidia, Meta and Oracle.
The $144 billion figure is the company’s worst-case scenario. It assumes margins improve to 20% over ten years and growth slows to 20% per year starting in 2028. Anthropic would then make $89 billion in after-tax operating profit in 2035, which is what Meta has made over the past twelve months. An intermediate scenario, with annual growth of 36%, gives $1.1 trillion. This is still 43% below the target.
No pricing power, company says
“In short, we do not believe Anthropic has a viable business.”
The report claims that open source models make it difficult for closed models to make a profit. It also states that Anthropic’s Claude Fable 5 and Mythos 5 cost approximately 18 times more than DeepSeek R1 for input tokens and 23 times more for output tokens.
The company says it has not seen Anthropic’s prospectus, which was filed confidentially, and that it worked from reports of financial leaks. Its track record in major listings is mixed, CNBC reported. WeWork withdrew its IPO six weeks after the company called it the most ridiculous of 2019. Its 2020 pick, DoorDash, held up as a public company. Anthropic did not respond to CNBC.
The same numbers, read differently
PitchBook senior analyst Harrison Rolfes analyzed the multiples on October 5. At $2 trillion, Anthropic would be valued at about 435 times its 2025 revenue, he writes. Compared to annualized revenue of more than $46 billion in the second quarter, the multiple falls below 44 times. Compared to the execution rate of more than $65 billion reached at the end of July, that’s a little less than 31 times.
Rolfes also noted that $34 billion of the $42 billion loss in 2025 came from a non-cash revaluation of convertible notes. The operating loss was more than $8 billion. He wrote that third-quarter revenue well above $16.3 billion would strengthen the case for valuation. The $16.8 billion Anthropic announced to investors this week crosses that mark. Rolfes added that this alone still wouldn’t justify $2 trillion.
Pricing an AI that could go rogue
The third question is one that most companies never ask investors. The draft filing devotes about 80 of its 261 pages to risk factors, Reuters reported on September 29. They warn that AI “may pose catastrophic or existential risks to humanity.” They also claim that models could exhibit “self-preservation behaviors”, including attempts to “resist stopping”. SpaceX devoted about 38 of its 277 pages to risk factors, Reuters noted.
New Constructs places this warning first on its list of red flags. Anthropic has already described such behavior in its own research. In May, he said Claude learned how to blackmail from stories about evil AI.
Bloomberg asked on October 8 how investors should assess this risk. Slow Ventures partner Sam Lessin has proposed that the US government force AI labs to put up $1 trillion in cash for any future cleanup.
“The liability could be a $2 or $4 trillion risk to the company, even before you consider the mass extinction that they’re talking about,” Lessin told Bloomberg.
Bloomberg also noted that the debate has not hurt SpaceX, which has focused much of its projects on AI. Its shares are trading more than 20% above their June IPO price.
Public buyers won’t have much say, which Rolfes says warrants a discount. The seven co-founders will retain 50.1% of the voting power after the listing via a Founder LLC.
Gn bussni
