Investors in Claude developer Anthropic are reportedly setting a $2 trillion target for its October IPO in an eye-watering and potentially risky play to set a record valuation.
According to the Financial Times, “half a dozen” Anthropic investors confirmed the target valuation, citing the company’s projections of $100 billion to $120 billion in revenue by the end of 2026. One claimed that $2 trillion was actually a lowball figure, given Anthropic’s rate of expansion, and suggested $3 trillion “on the incredibly low side” instead. (A totally unbiased figure, we’re sure.)
“It’s easy to come up with challenges,” one investor told FT. “But the company continues to be in first position in performance, positioning and what people want exposure to.”
Anthropic itself has yet to settle on a number, FT reported, and there’s conveniently no “publicly listed US peer” to compare the number against. So, this may be wishcasting. That said, it’s not that much higher than SpaceX’s $1.7 trillion IPO. While Elon Musk’s rocket company briefly shot to around $3 trillion, shares have been wildly variable since and are currently hovering around the IPO price of $135 per.
The company has had some hurdles as of late. In June, the White House freaked out about the release of frontier models Mythos 5 and Fable 5, forcing Anthropic to temporarily yank them from the public. Explanations have ranged from Fable 5 following a command to “fix this code” to a report that the models were capable of hacking into NSA and Cyber Command systems like it was nothing.
While Anthropic has since released Fable 5, its cybersecurity-focused Mythos model remains restricted to a handful of security firms participating in a project to find and patch software flaws using the model before it’s released. Anthropic [recently claimed](https://www.npr.org/2026/08/01/nx-s1-5914852/anthropic-openai-models-hack-cybersecurity) in a blog post that on three separate occasions, models undergoing testing (including Mythos) had broken out of their sandbox environments and launched attacks on real companies sharing the name of fictional targets.
Incidentally, its customers have complained about [too many guardrails](https://techcrunch.com/2026/06/10/cybersecurity-researchers-arent-happy-about-the-guardrails-on-anthropics-fable/).
Another challenge, FT reported, is cost. It cited data from AI analytics firm Artificial Analysis that show Anthropic’s models cost over 2.5x than rival OpenAI, and Chinese open-weight models are far cheaper to run than either. Building out AI infrastructure has also proven so expensive that the Federal Reserve has warned it could fuel inflation via its impacts on the electricity and computer hardware markets, CNBC recently reported. Companies have already begun to limit token use over cost concerns, and a recent Mavvrik/Benchmarkit poll showed as many as one in four have delayed or canceled AI projects due to cost.
Anthropic didn’t immediately respond to a request for comment on this story, but we’ll update if we hear back.