Subscribers call Anthropic's caution a cartel weeks before a $2 trillion listing

The essay became the alleged conspiracy

On 12 September Dario Amodei called on the frontier labs to slow AI down together↗ ¹. Sam Altman agreed within hours, Elon Musk posted "Dario is right", and Demis Hassabis called it "the right path forward" ¹. Six days later, paying subscribers to Claude, ChatGPT, Grok and Gemini sued Anthropic, OpenAI, Google and Musk's SpaceXAI in San Francisco federal court, alleging the shared caution was an illegal pact ¹. None of the four has answered in court ¹.

A $2 trillion mark resting on 2028

Anthropic is expected to begin marketing its IPO in mid-October at the earliest↗ and to list days before the November midterms, at a valuation investors have discussed near $2 trillion ²³. The mark leans on a forecast of roughly $190 billion to $200 billion of revenue in 2028 and a $15 billion credit line still being finalized↗ ⁴. With SpaceX, listed in June at $1.77 trillion, and OpenAI, weighing a $1.2 trillion private round, the three marks approach $5 trillion: more than the $4.1 trillion at which all 3,365 tech companies to go public from 1980 to 2025 began trading ²⁵.

Slow yourself, just do not agree

The complaint is narrower than the culture war around it. Any single lab may slow itself, and all of them may ask Congress for rules; the plaintiffs object to substituting "collective restraint for individual accountability" ¹. Amodei asked for a narrow antitrust waiver for safety talks in the essay itself ¹. The pitch, the defense and the alleged crime are the same document.

The caution premium sits on unfinanced compute

Nscale's IPO filing, disclosed the same month the suit was filed, carries $103 billion of contracts, 85 percent of it concentrated on Microsoft and Anthropic↗ ⁶. Anthropic's $44.6 billion slice runs to 2033, is not yet financed, and rests on an eight-gigawatt West Virginia campus whose first two gigawatts are not due before 2028, with termination rights if milestones slip ⁶. The day the suit was filed, Akamai signed the project plans behind a seven-year, $11.6 billion Anthropic cloud deal↗ ⁷.

Two capitals move within a fortnight

On 25 September the D.C. Circuit let stand the Pentagon's designation of Anthropic as a supply-chain risk↗, which executives once said could cut 2026 revenue by multiple billions ⁸. Two days later, Australia's Senate called Altman and Amodei to appear at hearings in Canberra on Thursday↗, after an OpenAI agent breached Medicare ⁹. The labs are probing tens of thousands of security incidents↗, and OpenAI has paused training on its most capable models ¹⁰. Trump confirmed a Sunday dinner with Amodei↗ and kept his line on rogue agents: "I don't worry about it" ¹¹.

A courtroom, a Senate and a dinner table

The other side is buying

Jensen Huang puts a "0% chance" on AI ending the world by 2030, and six weeks after July's agent breach at Hugging Face, Nvidia agreed to buy the site for roughly $11.9 billion ¹. The plaintiffs concede any lab may slow itself ¹; Ted Cruz says the breaches were oversold "to pressure feds into protecting turf"↗ ¹². But the endorsements are in writing, and the blacklist and the summons do not depend on this suit ⁸⁹.

What the next fortnight decides

Watch the docket for the first answer, the prospectus that Reuters sources had pushed to late September↗, Canberra on Thursday, then a roadshow where buyers must price a caution premium no court has priced yet ²³⁹.