Skip to article
From seven founders to a trillion-dollar listing A glowing line climbs from a cluster of seven small lights on the left, marking Anthropic's founding in 2021, through labelled milestones to a peak marked ANT, 19 November 2026. A stock ticker runs across the top. Below the peak, a row of data-centre blocks gives off heat haze and red pulses marking local opposition. ANT ▲ 97.17 · +17.1% · IPO 83.00 · OPEN 111.50 · HIGH 115.10 · LOW 92.72 · NASDAQ GLOBAL SELECT · 19 NOV 2026 ·    ANT ▲ 97.17 · +17.1% · IPO 83.00 · OPEN 111.50 · HIGH 115.10 · LOW 92.72 · NASDAQ GLOBAL SELECT · 19 NOV 2026 ·    ANT ▲ 97.17 · +17.1% · IPO 83.00 · OPEN 111.50 · HIGH 115.10 · LOW 92.72 · NASDAQ GLOBAL SELECT · 19 NOV 2026 ·    2021 · SEVEN LEAVE OPENAI MAR 2023 · CLAUDE SEP 2023 · SAFETY RULES FEB 2026 · $380B MAR 2026 · PENTAGON FIGHT 19 NOV 2026 · ANT 41 BALLOT MEASURES · 40 AGAINST
AI · Markets · A short history

The Careful Company Goes Public


Seven researchers left OpenAI in 2021 because they feared what AI could do. On 19 November 2026 their company floated at a value of more than a trillion dollars, while protest signs outside its headquarters warned about jobs, extinction risk and data centres.

Dario and Daniela Amodei stood on the floor of the Nasdaq MarketSite in New York on the morning of 19 November 2026, beside the board of the company they had started out of a shared sense of alarm five years earlier.↗ Shares of Anthropic, ticker ANT, had been priced the night before at $83. When they finally changed hands at about 11:48 a.m., the first trade was $111.50.↗ By the close they had settled at $97.17, up 17%, and the company that once described itself as a safety lab was worth about $1.76 trillion on a fully diluted basis.↗ Back in San Francisco, protest signs had been left in front of its headquarters.↗

Inside the company, Bloomberg reported, it was "largely business as usual," with no splashy celebrations planned.↗ That restraint fits the story Anthropic has always told about itself. It also sits strangely next to the numbers. The offering raised a record $81.3 billion, more than SpaceX's $75 billion in June, and another $26 billion came from technology companies buying shares on the side.↗ This is the short history of how a group of worried researchers ended up running the largest stock-market debut on record, in a year when most Americans said AI worried them more than it excited them.↗

DisclosureThis article was researched and drafted with Claude, which is made by Anthropic, the company it describes. That's a conflict of interest, so the critics' case below is given in their own words and every figure is linked to a source outside the company wherever one exists.

I · San Francisco, December 2020

A resignation, and six more


Dario Amodei had joined OpenAI in 2016 and risen to vice president of research, close to the centre of the decisions that produced the GPT models. His departure was announced on 29 December 2020. He didn't leave alone: his sister Daniela and a group of colleagues walked out with him.↗ Seven of them became Anthropic's co-founders, among them Tom Brown, lead author of the GPT-3 paper, interpretability pioneer Chris Olah and former technology journalist Jack Clark.↗↗

The founders' public explanation has barely changed in five years. They disagreed with OpenAI about how directly safety research should shape what gets built and released.↗ Later reporting described a far more personal break, including conditions Dario reportedly set for staying that OpenAI turned down.↗ Either way, the rivalry outlived the split. It's still running, and it now has a share price.

The new company was set up in early 2021 as a public benefit corporation, a legal form that lets directors weigh a stated mission alongside profit.↗ Its stated aim, as reported that June, was to build "reliable, interpretable, and steerable AI systems."↗ In plain terms: systems that behave the same way twice, whose reasoning people can inspect, and that do what they're told.

Here's the bet that shaped everything after. The founders believed that you can't learn to make powerful AI safe without working on powerful AI. So a company created out of worry about the frontier decided it had to stand on the frontier. That logic explains the papers on interpretability. It also explains the data centres, the trillion-dollar valuation and much of the anger now aimed at the company.

If you were ten Imagine some people who think fireworks are dangerous. They could stay away and complain. Or they could open the biggest fireworks factory in town, so they can study what makes fireworks explode and set the safety rules everyone else copies. Anthropic chose the factory. The trouble is that a factory needs money, land and power, and the neighbours start to have opinions.
II · 2023, a chatbot and a rulebook

Claude arrives, the rules follow


For its first two years Anthropic was mostly a research lab. The public timeline begins in March 2023, when it released two versions of its assistant: Claude, and a lighter, faster Claude Instant.↗ It earned its first dollar of revenue around then. By February 2026 it said that run-rate revenue had grown more than tenfold in each of the three years since.↗

September 2023 brought two things that still define the company. The first was its Responsible Scaling Policy, published on 19 September: a public commitment not to train or deploy models capable of causing catastrophic harm unless matching safety and security measures were in place.↗ The policy has been revised many times since, most recently in 2026.↗ The second was a governance device called the Long-Term Benefit Trust: five members with no financial stake in the company, given the power to choose a growing share of its board, eventually a majority.↗

The same autumn, the money arrived. Amazon announced an investment of up to $4 billion in September 2023, and Google committed $2 billion the following month.↗ Those cheques bought cloud computing as much as equity. They also tied the safety lab to two of the largest companies on Earth, which would later become its biggest suppliers and its competitors at the same time.↗

  1. Dario Amodei's departure from OpenAI is announced. Daniela Amodei and others follow.
  2. Anthropic is founded by seven former OpenAI staff as a public benefit corporation.
  3. Claude and Claude Instant are released.
  4. Responsible Scaling Policy and Long-Term Benefit Trust published. Amazon pledges up to $4 billion; Google adds $2 billion in October.
  5. Dario Amodei warns AI could wipe out half of entry-level white-collar jobs.
  6. $13 billion raised at a $183 billion valuation.
  7. $30 billion raised at $380 billion. Run-rate revenue: $14 billion.
  8. The Pentagon labels Anthropic a "supply chain risk." Anthropic sues.
  9. Claude Mythos Preview is released only to a small group of trusted organisations.
  10. Confidential IPO filing.
  11. Public prospectus filed: Nasdaq, ticker ANT.
  12. First day of trading. Closes at $97.17.

Fig. 1 · Five years from walkout to listing. Sources linked in the text and listed below.

III · 2025 to 2026, money and fights

A warning, a blacklist, a filing


In May 2025, during the company's first developer conference, Dario Amodei gave Axios the interview that made him the most quoted pessimist in the industry he leads. AI, he said, could wipe out half of all entry-level white-collar jobs and push US unemployment to 10–20% within one to five years. He told AI companies and governments to stop "sugar-coating" it.↗ He also admitted the irony of selling the products he was warning about.↗

Investors didn't hear a warning. They heard demand. In September 2025 Anthropic raised $13 billion at a $183 billion valuation. Five months later it raised $30 billion at $380 billion, with run-rate revenue of $14 billion and its coding tool, Claude Code, already bringing in more than $2.5 billion a year.↗

Then came the fight that made Claude briefly the most downloaded iPhone app. In February 2026, talks over a $200 million Pentagon contract broke down.↗ Anthropic wouldn't drop its limits on using its models for surveillance or autonomous weapons, and Defense Secretary Pete Hegseth had the company designated a "supply chain risk," a label usually reserved for firms tied to foreign adversaries.↗ Anthropic sued on 9 March, calling the government's actions "unprecedented and unlawful."↗ Claude shot to the top of the App Store.↗ The legal battle was still running on the eve of the IPO.↗

In April came Claude Mythos Preview, a model the company held back from general release while it built extra cybersecurity safeguards, offering it first to a small group of trusted organisations.↗ (This site covered it in The Model They Won't Release.) And on 1 June, Anthropic confidentially filed to go public, proposing a valuation of $965 billion.↗

Anthropic's valuation, September 2025 to November 2026 Horizontal bars: $183 billion in September 2025, $380 billion in February 2026, $965 billion proposed in June 2026, about $1.49 trillion at the IPO price on 18 November 2026, and $1.76 trillion fully diluted at the first-day close on 19 November 2026. SEP 2025 · PRIVATE ROUND FEB 2026 · PRIVATE ROUND JUN 2026 · PROPOSED IN FILING 18 NOV · IPO PRICE ($83) 19 NOV · FIRST-DAY CLOSE $183B $380B $965B ~$1.49T $1.76T
Fig. 2 · Bars to scale. The June figure was a proposal, not a trade (dashed). The 19 Nov figure is fully diluted. Sources: Reuters via AOL, Shacknews, Financial Times via The Next Web, CNN.

The public prospectus landed on 26 October: 318 pages, listing on the Nasdaq under ANT.↗ It showed $19.7 billion of revenue in the twelve months to 30 June 2026,↗ $16.3 billion of it in the first half of 2026 alone.↗ A later amendment put third-quarter revenue at nearly $17 billion.↗ It also showed what that growth costs. Anthropic reported a first-half net loss of $80.7 billion, though $72.4 billion of that was a non-cash accounting adjustment rather than money spent.↗

The governance looked like nothing else on Wall Street. The co-founders would hold 50.1% of the voting power after the listing,↗ and the Long-Term Benefit Trust could elect or remove a majority of the board.↗ The prospectus asked buyers to try "not to think of your purchase in purely transactional terms," and warned that profits would be balanced against "the public good."↗ Few companies have told shareholders, in writing, that they may come second. Demand still reached close to seven times the shares on offer, and the price was lifted above its $70–$78 range to $83.↗

ANT's first trading day, 19 November 2026 A price range chart from 80 to 120 dollars. A dashed line marks the IPO price of 83 dollars. A single candle shows the open at 111.50, the close at 97.17, the day's high at 115.10 and its low at 92.72. Every price traded was above the IPO price. $80$90$100$110$120 IPO PRICE $83.00 (SET 18 NOV) HIGH $115.10 OPEN $111.50 · FIRST TRADE 11:48 AM ET CLOSE $97.17 · +17.1% ON IPO PRICE LOW $92.72
Fig. 3 · A big pop that faded, but never below the offer price. The candle is red because the stock closed below where it opened. Sources: CNN; Nasdaq data via Wolf Financial.

To understand why the company needed so much money, look at its obligations. In October the prospectus put committed spending on future computing and infrastructure at $542–$567 billion.↗ An amendment on 2 November raised its non-cancellable purchase commitments and leases to $895.4 billion, most of it for compute, with more than 3.5 gigawatts of capacity expected online by the end of 2026.↗ Substantially all of that spending is expected to fall between 2031 and 2045.↗

Revenue compared with spending commitments Two bars to scale. Twelve months of revenue to 30 June 2026, 19.7 billion dollars, is a sliver. Purchase commitments and leases disclosed on 2 November 2026, 895.4 billion dollars, fill the width of the chart. REVENUE, 12 MONTHS TO 30 JUN 2026 PURCHASE COMMITMENTS AND LEASES, AS OF 2 NOV 2026 $19.7B $895.4B, MOSTLY COMPUTE AND INFRASTRUCTURE
Fig. 4 · Bars to scale. The commitments run to 2045, so this isn't one year's bill, but it shows why Anthropic wanted the public markets. Sources: Renaissance Capital; amended S-1 via Crypto Briefing.
IV · New York and San Francisco, November 2026

Three fears at the door


The bell rang into a mood that was nothing like the dot-com era. The Nasdaq 100 was 13% below its early-October peak when the offer was priced,↗ and the public had soured. Anthropic itself named the backlash against AI and data centres as a risk factor for investors.↗ Three fears came up again and again.

The first fear: jobs

In June 2026, 71% of American adults told Pew they expected AI to mean fewer jobs over the next twenty years, up from 64% in 2024. Only 5% expected more.↗ It isn't just an American worry. Across 37 countries, the median share expecting fewer jobs was 46%, against 9% expecting more.↗ The awkward part for Anthropic is that its own chief executive is one of the loudest voices behind the fear. Not every economist agrees with him: Apollo's chief economist, Torsten Sløk, argued in June that there was still no evidence AI was killing jobs overall.↗ (This site looked at the plans for sharing the gains in The Last Payslip.)

The second fear: losing control

The sharpest attack came from people who share Anthropic's view that advanced AI could be catastrophic. On the eve of the listing, at least 80 researchers, executives and academics, including Turing Award winner Judea Pearl, physicist Max Tegmark and Tristan Harris of the Center for Humane Technology, had signed what they called the "misanthropic statement." It said: "Anthropic is not a safe AI company and we must reject its reckless attempt to amass power in the name of the greater good." Tegmark accused the company of warning in public about AI that improves itself while being "pedal to the metal on RSI" (recursive self-improvement) in private.↗ Amodei had already pushed back in August against the charge that his own warnings were feeding the backlash.↗ This site has tracked how the builders themselves rate the danger in p(doom).

The third fear: the buildings

The most concrete opposition was local. On 3 November, voters in ten US states decided 41 ballot measures about data centres and sided with restrictions in 40 of them, according to CNBC's analysis. The one exception was Harwood, North Dakota.↗↗ Residents near new projects asked who actually benefits once the construction crews leave.↗ Anthropic's filing describes on-site natural gas generation as part of how it plans to power new capacity,↗ which sits uneasily with climate goals. In its favour, the company pledged in February to cover any rise in consumer electricity prices caused by its own data centres. The Register pointed out the catch: the pledge covers only the sites Anthropic builds itself, not the capacity it leases from Amazon, Google and Microsoft.↗

Those fears met in the street. A coalition called Stop Funding Billionaires, made up of labour, student and community groups joined by climate activists, protested at Anthropic's New York office under the banner "Block the IPO." New York City's comptroller, Mark Levine, had asked the company in October to pause the offering.↗ The coalition's petition asked banks and fund managers for "a moratorium on investments in AI and data centers."↗

The three fears, side by side
FearWhat critics sayWhat the company says or didOne number
JobsAI labs are automating away the first rung of white-collar careers.Its CEO says the risk is real and governments should prepare.71% of US adults expect fewer jobs
Control"Not a safe AI company": racing toward systems that improve themselves.Publishes a scaling policy; a trust can remove a majority of the board.80+ signatories to the "misanthropic statement"
EnvironmentData centres strain power grids and the towns around them.Pledged to cover consumer power-price rises from its own sites; plans on-site gas.40 of 41 local ballot measures went against data centres

Sources: Pew Research Center; The National; The Register; CNBC via Latin Times; Anthropic's S-1 as reported.

$97.17first-day close, 17% above the $83 IPO price
$895.4Bin purchase commitments and leases, mostly for computing
40 of 41data-centre ballot measures decided in favour of restrictions on 3 November

So who is right? Both sides, partly, and that's what makes this listing historic rather than merely large. The critics are right that a company built on fear of the race now has to keep winning it, with public shareholders watching every quarter. A Truist study of 32 of the biggest IPOs found more than half saw their shares fall at least 50% at some point in their first year.↗ That's the pressure that will test the trust, the scaling policy and the founders' 50.1%. Anthropic's defenders are right that no other company of this size has written its safety commitments, its mission and its critics into a prospectus and still found buyers at seven times the shares available.

Yahoo Finance's debut-day report quoted Dario Amodei describing the moment in terms few chief executives would use for a stock listing: "The advent of machine intelligence is one of the seminal moments in human history."↗ Five years earlier he'd walked out of OpenAI because he thought that moment needed more caution. Now the people he has to convince include everyone who owns a share of ANT, and everyone who lives next to the buildings it pays for.

What to watch

  • First resultsAnthropic's first quarterly report as a listed company, and whether revenue keeps pace with its compute bills.
  • The courtsThe outcome of its lawsuit over the Pentagon's "supply chain risk" label.
  • OpenAIIts rival is preparing its own listing. Two public AI labs will be compared quarter by quarter.
  • Data centresWhether local opposition delays the 3.5 gigawatts Anthropic expects online, and how much of it runs on gas.
Ko-fi Buy me a coffee
Scroll to Top