The most valuable IPO ever attempted opened its prospectus by warning that its own product could pose existential risks to humanity. Anthropic made its S-1 public on Monday, confirming a Nasdaq listing under the ticker "ANT," with a debut expected around Thanksgiving and a valuation bankers have put at $1.5 trillion to $2 trillion. The filing shows growth that is hard to find a precedent for: $16.3 billion in revenue in the first half of 2026, up from $1.2 billion in the same period a year earlier. It also shows a company still deep in the red, with risk factors that run from a feud with the Pentagon to CEO Dario Amodei's personal reputation to the possibility that advanced AI ends badly for everyone. Anthropic could raise as much as $100 billion.
That combination — record growth, unproven profitability, and apocalyptic hedging in the same document — is the week's defining tension, and it put the Amodei siblings on the move. Dario and Daniela Amodei both rose on the index. The filing reaffirmed a pledge the seven cofounders — Dario Amodei, Daniela Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish and Chris Olah — first made in a January essay by Dario: to give away 80 percent of their personal equity. At the latest private price their combined stakes are worth roughly $99 billion, making the pledge worth close to $80 billion, or about $11 billion a founder on average. The catch, which Fortune flagged, is that the S-1 calls the pledge voluntary and revocable rather than binding, and a Founder LLC keeps the seven holding 50.1 percent of the voting power even if they give the shares away. "The thing to worry about is a level of wealth concentration that will break society," Dario wrote in that January essay — an unusual line to find behind a nine-figure payday. Daniela Amodei, Anthropic's president, separately drew headlines for taking aim at OpenAI over a Pentagon deal and questioning its rival's motives — a reminder that the offering lands amid open friction with Washington and with the competitor Anthropic most defines itself against.
Atop the index, though, sits someone who said nothing. Liang Wenfeng held his number-one spot on the strength of money, not words: his startup DeepSeek is wrapping up a round of about $15 billion, or roughly 100 billion yuan, which Bloomberg called one of the largest in China's history. The target was lifted at least twice — from 50 billion to 80 billion to about 100 billion yuan — with battery giant CATL and Tencent among the biggest backers and a pre-money valuation near $74 billion. The timing was pointed: the raise came together roughly two days after Washington ordered federal civilian agencies to pull DeepSeek's software over security concerns. Liang told investors in September that annualized revenue had reached a $1 billion run rate, more than double the $400 million to $500 million of July, and the company is said to be accelerating a Shanghai listing for early 2027 — potentially beating OpenAI to public markets.
The money story had a counterpoint in the timeline story. Demis Hassabis, holding steady, again pulled in his AGI forecast, now putting human-level machine intelligence two to three years out, down from the three-to-five he was giving earlier this year — while warning regulators not to rush rules before the technology lands. Jensen Huang, also steady, spent the week reinforcing scarcity as a business model: Nvidia is effectively sold out, with the chief executive of Palo Alto Networks relaying that the company can't commit chips for 2028, even as reports surfaced that Nvidia wants roughly $2,000 per GPU in annual software fees and backers lined up a $2 billion AI hub in Kazakhstan.
Among the risers, Clément Delangue of Hugging Face broke out by picking a fight, calling a new Netflix documentary about an AI agent attack "super disappointing" and saying it went "full doomer mode," while telling a Miami summit that every company will soon become an AI builder. The notable faller was Jeff Dean — not for anything he did, but because the index caught noise around his name, from soccer scores to an unrelated obituary, with only an unverified report that a 17-year-old researcher had left Moonshot's Kimi for a Dean-linked venture standing out as possible signal.
The throughline: in a single week, the field's two most consequential labs — one American and closed, one Chinese and open — each locked in a war chest, while the people building the models kept warning, in prospectuses and on stage, that the thing they're racing to build might be dangerous. The capital has decided. The economics, and the safety case, have not.
Anthropic (an-THROP-ik) just filed to go public, and it could be the biggest IPO the market has ever seen — but here's the strange part: the company spends its own prospectus warning that the technology it sells could pose existential risks to humanity. That filing went public on Monday. It confirms a listing on the Nasdaq under the ticker A-N-T, a debut expected around Thanksgiving, and a valuation that bankers are putting somewhere between one-and-a-half and two trillion dollars.
And the growth is genuinely hard to wrap your head around. Anthropic brought in sixteen-point-three billion dollars in revenue in just the first half of this year. A year earlier, that same six months was one-point-two billion. So that's more than a tenfold jump. The company could raise as much as a hundred billion dollars in the offering. But it's still losing money, and its list of risk factors is wild — it runs from a feud with the Pentagon, to the personal reputation of its CEO, Dario Amodei (DAR-ee-oh uh-MOH-day), all the way to the possibility that advanced AI ends badly for all of us.
That tension — explosive growth, no proven profit, and doomsday language all in the same document — is really the story of the week, and it pushed both Amodei siblings up the index. Dario and his sister, Daniela Amodei (dan-ee-EL-uh uh-MOH-day), who's the company's president, are both rising. The filing restates a promise the seven cofounders made earlier this year: to give away eighty percent of their personal stock. At today's private valuation, their combined stakes are worth around ninety-nine billion dollars — so that pledge is worth close to eighty billion, roughly eleven billion per founder on average. The catch? The filing calls that pledge voluntary and revocable — not legally binding. And through a founder holding company, those seven still keep just over fifty percent of the voting power, even if they hand the shares away. Dario himself wrote, and I'm quoting, "The thing to worry about is a level of wealth concentration that will break society." Which is a remarkable thing to say right before one of the largest paydays in history. Daniela, for her part, also made news taking a shot at OpenAI over a Pentagon deal, questioning her rival's motives.
Now, the person actually sitting at number one on the index this week didn't say a word. That's Liang Wenfeng (lee-AHNG wun-FUNG), and he's there purely on money. His company, DeepSeek (DEEP-seek), is closing a round of about fifteen billion dollars — roughly a hundred billion yuan (yoo-AHN) — which Bloomberg called one of the biggest in China's history. They raised the target at least twice to get there. Backers include the battery giant C-A-T-L and the tech company Tencent (TEN-cent). And the timing was pointed: this came together about two days after Washington ordered federal civilian agencies to pull DeepSeek's software over security concerns. Liang told investors back in September that the company had hit a one-billion-dollar annual revenue run rate — more than double what it was in July — and DeepSeek is reportedly speeding up a Shanghai (shang-HYE) listing for early 2027, which could put it on public markets before OpenAI.
A quick word on the timeline debate. Demis Hassabis (DEM-iss huh-SAH-biss) of Google DeepMind (DEEP-mynd) pulled in his forecast again — he now thinks human-level AI is two to three years away, down from the three-to-five he was saying earlier this year — and in the same breath, he told regulators not to rush the rules before the technology actually arrives.
Meanwhile Jensen Huang (JEN-sun HWAHNG) spent the week leaning into scarcity as a strategy. Nvidia (en-VID-ee-uh) is basically sold out — the head of Palo Alto Networks (PAL-oh AL-toh) said Nvidia told him it can't even commit chips for 2028 — and there are reports Nvidia wants around two thousand dollars per GPU in annual software fees, plus a new two-billion-dollar AI hub being lined up in Kazakhstan (kah-zak-STAN).
Among the climbers, Clément Delangue (kleh-MON duh-LAHNG), who runs Hugging Face (HUG-ing face), broke out by picking a fight — he called a new Netflix documentary about an AI agent attack "super disappointing" and said it went, quote, "full doomer mode." And the big faller was Jeff Dean — though honestly not for anything he did. The index just caught noise around his name, from soccer scores to an unrelated obituary, with only an unverified report about a teenage researcher leaving Moonshot's Kimi (KEE-mee) for a Dean-linked startup standing out as anything real.
So here's the thread tying it all together. In one week, the two most important labs in the world — one American and closed, one Chinese and open — each locked in an enormous pile of cash. And the very people building these systems kept warning, in their filings and up on stage, that the thing they're racing to build might be dangerous. The capital has clearly made up its mind. The economics, and the safety case, still haven't.