Anthropic's $2 trillion listing would price it at 31 times the revenue it reported in July

Thirty-one times revenue is what the target implies: $2 trillion of valuation against the $65 billion Anthropic told shareholders it booked in July ^[1][5][9]. The 6,000 customers the company counts at $100,000 a year or more come to at least $600 million annually ^[6][7], which is under 1% of that July figure ^[8]. Against the customer disclosure alone, $2 trillion is more than 3,000 times ^[24]. Both numbers came from the same company.

The profit line is narrower than it first looks. Anthropic said adjusted operating income will stay positive for a third quarter in a row on gross margins above 80% ^[3], and the account states those figures come before model-training and revenue-sharing costs ^[4].

On the demand side, Cryptopolitan reports that Nvidia, Palantir and Booz Allen Hamilton have begun reducing or eliminating their use of Anthropic and OpenAI models, on concerns that those systems could learn from customer intellectual property ^[10]. Both labs deny training on information from enterprise users while retaining "some metadata" ^[11]. Microsoft is selling directly into that doubt, offering isolated cloud environments and models run on privately managed servers with no transfer of user data outside ^[12]. If those three buyers sit in the six-figure cohort, each exit is at least $100,000 a year ^[6], and the bigger question is how many of the 6,000 follow them.

The political pressure is specific and named. White House AI adviser David Sacks called for the IPO to be paused until whistleblower claims can be investigated, after former Anthropic pretraining researcher Jacob Coxon resigned and accused Anthropic and OpenAI of racing toward self-improving superintelligence ^[16][17]. Senators Ted Cruz and Bernie Sanders have separately pressed for tighter safeguards ^[18]. Anthropic's own chief executive, Dario Amodei, said on September 12 that the industry must pace frontier models because "building it too fast is reckless" ^[15]. Microsoft's Mustafa Suleyman, describing the consultation on his company's provisional code for its MAI models, said: "We got feedback from people that they wanted to see even more explicit commitment to AI always working in the service of people and not trying to replace them" ^[13][14].

In my view the safety argument moves the price and leaves the timetable where it is. Anthropic's stated reason for listing, according to the account, is that being public makes safety easier to demonstrate through transparency ^[20], and OpenAI's calendar sits a year behind at 2027 ^[19]. Two other readings are live. If the restrictions at Nvidia, Palantir and Booz Allen are confined to workloads that touch customer IP, they cost Anthropic a handful of line items and leave the six-figure cohort intact. If a second official voice joins Sacks, the 2026 date is what slips, and the $1.77 trillion SpaceX reached in June stays the high-water mark for a while ^[2][16].

What would show me wrong: the customer count coming in below 6,000 at the next shareholder update, or a prospectus restating that July revenue figure on a different basis ^[5][6].