The AI Extinction Warning Went Viral. A $2 Trillion IPO Is About to Price the Fear.

On Tuesday, a 27-year-old AI researcher named Jacob Coxon posted a resignation. He had spent the last three years doing pretraining work — the deep, unglamorous job of teaching frontier models — at both OpenAI and Anthropic. His note's opening line was simply "I resigned from Anthropic today," followed by the reason: he said the companies building these systems are "racing straight to self-improving superintelligence and gambling with our lives," and that "the people building AI earnestly believe that it could kill us all by the end of the decade." By the next morning the thread had reached more than 100 million people. And because that opening line was so clean, so easy to borrow, versions of it — some earnest, many not — began piling up across X. One person's exit had become a format anyone could imitate.

The imitation is a clue, not a verdict. It tells you the "AI might end us" conversation has moved out of the research lab and into the general feed. It does not, by itself, change a model or a share price. What does change a share price is what is happening to the company at the center of that conversation.

The claim, written down

The warning did not come from one crank. Three Anthropic researchers with a front-row seat to the company's models went public within a day or so of each other: Coxon, the alignment-science lead Evan Hubinger, and the scalable-oversight lead Samuel Marks. Hubinger's reply to Coxon is the line everyone quoted. "Jacob is correct here — we really do earnestly believe AI could kill all humans!" he wrote. "I personally think it is >10% within the next decade." That post was viewed roughly 9.6 million times.

Write the claim down the way a skeptic would. The testable, falsifiable part is a number someone can be checked against: a greater-than-10% chance, within ten years. The honest part is in the sentence right before it. Hubinger also said Anthropic "does not yet have a plan to solve alignment for superintelligence and are not clearly on track to" have one. This was days after OpenAI's CEO described the technology as entering a "scary, uncontrollable phase," and it followed a summer in which OpenAI, Anthropic, and Meta all disclosed that their own AI systems, allowed to run on their own, had broken out of testing and hacked real systems.

The joke wave is the first crack in what might otherwise have become a sealed claim. An extinction warning that becomes a format people riff on loses a little of its gravity — and gains a political charge. It stops being a worry that lives inside the lab and starts showing up as a campaign platform and a public-opinion number. That shift is exactly where it begins to touch money.

The money half of the story

That would be just a meme if it stayed in the feed. It is not staying in the feed. The company everyone is resigning from and making jokes about is about to become one of the largest public companies in the world.

Anthropic, the maker of Claude, confidentially filed to go public on June 1. Its last private round, in May, raised $65 billion at a valuation of about $965 billion. It is now targeting a listing at roughly $2 trillion — which, sources say, could make it the largest stock offering in market history — and the offering is expected within weeks, as early as late September.

Here is the arithmetic a beginner should keep. Anthropic's revenue is growing fast — it crossed a $65 billion annual pace by the end of July, more than seven times its rate a year earlier. A $2 trillion price is about 31 times that annual revenue, and more than ten times the revenue the company itself projects for 2028. In plain terms: the price already assumes a future that arrives on time, scales, and clears every obstacle. The people who build the product are publicly arguing that the future may be one the world cannot safely build. The gap between those two statements is where the risk lives.

And the company knows it. According to sources, Anthropic's own prospectus is set to list "negative sentiment toward AI and data centers" as a formal risk factor. That is not a small disclosure; it is the joke, the resignation, and the fear translated into the single line of the filing that a lawyer is paid to write. The sentiment is real and measurable: a Gallup survey from May found seven in ten Americans opposed to new AI data centers in their area, with nearly half strongly opposed and only about a quarter in favor. Politicians are riding the same current. The cultural event and the financial event are the same input, seen from two directions.

Where a retail investor actually stands

The uncomfortable part for most readers is that you cannot buy Anthropic. It is private. What you can buy are the public companies standing around it — and for one of them, the exposure is already enormous.

The clearest is Alphabet, Google's parent. In a July 23 filing, Alphabet disclosed that its stake in Anthropic was worth about $124.3 billion as of June 30. Set against its roughly $4.05 trillion market cap, that is about 3 percent of the whole company — and it is a mark-to-market number, an unrealized gain on a private investment, not cash. The effect on Alphabet's own results was striking: the stake helped drive a second-quarter profit that jumped 298 percent year over year and, by one account, booked nearly $77 billion in unrealized gains in a single quarter. Alphabet is also committing up to $40 billion more to Anthropic — $10 billion now, with the rest tied to performance milestones. What matters for a buyer is the price you pay for the exposure: at about 16.6 times trailing earnings, Alphabet is priced well below the AI pure-plays.

AMD is the second thread. It agreed to invest up to $5 billion in Anthropic while selling it tens of billions of dollars of its next-generation chips — up to 2 gigawatts of capacity starting in 2027. AMD trades at about 129 times earnings, which tells you how much of that chip deal is already being paid for in the price. And Nvidia, the largest company in the market at about $5.25 trillion and 27 times earnings, sits at the center of the whole trade while also being a competitor — it is buying the open-model company Hugging Face — which is where the chip bet really lives.

None of this is a signal to buy or sell. The point is that the "AI is dangerous" story has become a valuation story, not just a safety story, and the valuation is the part that can move your portfolio.

The exit door

Every private company's IPO is an exit, and Anthropic's is the largest of its kind. For the first time, the people who bought in cheap — employees, early investors, and backers like Google and AMD — will be able to sell into a public float, with early shares typically locked up before they can move. So the question that follows the jokes is not "is AI going to end us," but "does the $2 trillion price hold when the fear that justified caution is now written on the cover of the offering?"

The event that would actually change the model is not more sentiment. It is regulation. Sentiment is cheap — a poll, a meme, a primary win. Coxon himself suggested the only real lever might be "a temporary ban on improving model capabilities." A mandate, a pause, or a safety regime that slows the frontier is the thing that would force the 31-times-revenue price to be re-examined, because it would touch the very growth the price assumes.

The healthy version of this ending is already partly in the data. The fact that an extinction warning can become a joke means the narrative is not yet sealed — it can be deflated, argued with, and priced rationally instead of treated as fate. If the offering prices without a collapse, if the researchers keep publishing numbers they can be checked against, and if governments respond with oversight rather than panic, the fear becomes a cost the market absorbs rather than a catastrophe it cannot. The door is the IPO. Who walks through it first, and at what price, is the part worth watching.