Anthropic's IPO: Record Debut Puts It Above OpenAI

An episode of Dan's AI Intel

Anthropic's debut gave AI its first public price for a pure frontier lab: investors believe the revenue and are still pricing the compute bill.

Published · By Dan Walter

Executive summary

Anthropic's Nasdaq debut gave the AI race something it has never had: a live public price for a pure frontier lab. By Thursday's close the market valued the maker of Claude at about $1.76 trillion, above the latest private price put on OpenAI. From now on the contest between the labs is scored every trading day, in public, and capital will follow whichever lab's numbers hold up.

The first day also showed what that price rests on. Investors chased the stock at the open, then sold half of its first-trade gain away on a day the whole market fell. The market believes the revenue, which reached a run-rate of about $81 billion in October. What it is still pricing is the bill: about $900 billion of signed compute and infrastructure commitments.

Previously on the show

Our S-1 episode, Anthropic's S-1: The AI Race Is Now a Balance-Sheet Race (episode 78, 2 November 2026), made two points. Serving AI already pays for itself at Anthropic. And compute it cannot cancel turns the race into a contest of capital. Back in August, The AI Bubble Is Real — But Only in One Layer (episode 49) argued that the exposed layer is the pure-model labs.

What has changed since: the filing became a price. Seventeen days after episode 78, the market has put a number on the balance-sheet race, and on the one layer episode 49 called exposed.

Why one share price matters to the AI revolution

Until Thursday, every frontier lab's value was set in private rounds by a small group of investors, many of them suppliers. Yorkville's Dan Ives put the change plainly: "For the first time a US frontier lab carries a standalone public market price, uncluttered by an ad business or a rocket business." That price now feeds pension funds, index funds and the cost of capital for everyone building AI.

A record raise, priced above the range

What happened at the IPO, in one line: Anthropic priced at $83 a share on the evening of 18 November, above a marketed range of $70 to $78, sold 980 million shares and raised $81.3 billion, the largest IPO on record. It began with 825 million shares and upsized after nearly $500 billion of non-binding orders, about $400 billion from institutions and close to $100 billion from individuals. The previous record was SpaceX in June, at $75 billion before its overallotment option.

Around the deal sat a $26 billion side placement at the same price, from Google, Nvidia, Microsoft, AMD, Broadcom, Micron and TSMC. Up to 20% of shares were set aside for individual investors. Amazon and Google already held stakes worth over $100 billion each.

Day one: a 34% pop, then half of it handed back

Trading before and after the opening bell told one story: a pop that buyers expected and got, then a fade toward a price the market would hold. Before the opening trade, grey-market and perpetual-futures prices implied a gain of about 25 to 27%. Dario and Daniela Amodei rang Nasdaq's opening bell, while protesters outside unfurled banners reading "freeze AI".

Exhibit — Anthropic opened 34% above its $83 IPO price and closed 17% up. Source: CNBC and AP (offer price, 18 Nov 2026); NBC News and Yahoo Finance live blogs (open, high, close, 19 Nov 2026). Compiled by Dan's AI Intel.

ANT first traded just before noon New York time at $111.50 and touched $115.10. Then it fell about 13% in thirty minutes, the worst start for a big IPO since 2008 by Yahoo Finance's count, and closed at $97.17, up 17%. That close came on a day the S&P 500 fell 2.02% and the Nasdaq 2.25%, as worries about Nvidia's China export problems spread across AI stocks. Nvidia lost 3.8% and SpaceX 4.8%. A 17% gain into a falling market is strength; a fade from the open is the market declining to pay any price for the story.

Individual investors traded hard on both sides. Vanda Research counted $255 million of retail turnover in the first half hour, more than double SpaceX's debut, but only $18.6 million of net buying. "Retail are trading the stock hard... but taking both sides rather than a one-way bid," said Vanda's Viraj Patel. One 20-year-old student asked for six shares at $83, was given one, and bought four more at about $97.

What people are saying: they believe the revenue and worry about the bill

The commentary splits into two camps: believers who say the revenue justifies the price, and sceptics who say slowing growth and the compute bill do not.

The believers point to growth. Revenue for the nine months to September was $33.2 billion, up about 1,200%. Ives called the debut "a big near-term win for the tech sector". Walser Wealth Management's Rebecca Walser called it "the show me moment for the AI story." Needham started coverage at buy with a $150 target; across five buy ratings the average target was about $116.

The sceptics point to the second derivative: revenue growth over the previous quarter fell from 144% in the second quarter to 47% in the third.

Exhibit — Anthropic's quarterly revenue growth slowed from 144% to 47%. Source: Bloomberg, as reported by The Next Web, 19 Nov 2026. Compiled by Dan's AI Intel.

Compute costs reached 58% of revenue last quarter. The net loss for nine months was $104 billion, mostly accounting charges on convertible notes. Plancorp's Callie Cox warned that hot IPOs rarely perform straight out of the gate. Jim Prosser, a former Google communications leader, warned of the opposite risk: "The second-worst thing that can happen to you is a wildly successful IPO because what that does is it changes the scorecard underneath you." Beyond the numbers, a "MisAnthropic Statement" signed by scientists the day before argued Anthropic is not a safe AI company. One former employee told The Verge the pace of AI development feels "untenable under normal market structures".

OpenAI and the rest of the field now have a price to beat

For OpenAI, the debut sets a public price to beat: Anthropic closed at about $1.76 trillion fully diluted, above OpenAI's reported $1.25 trillion pre-money mark, ahead of an OpenAI listing now aimed at 2027.

Exhibit — Markets valued Anthropic above OpenAI's latest private price on day one. Source: NBC News (close, 19 Nov 2026); CNBC (IPO price, 18 Nov 2026); Tech Funding News citing CNBC (OpenAI, 13 Nov 2026). Compiled by Dan's AI Intel.

The comparison is not like for like, a traded price against a private round, but it is the one investors will make. OpenAI's round was reportedly half the $30 billion it first sought, and its third-quarter revenue was put at about $10 billion against Anthropic's $17 billion, though the two count cloud-partner revenue differently. Sam Altman said at DealBook: "We plan to be public at some point." Willkie Farr's Eddie Best was blunter: "It's going to help OpenAI become public." The help cuts both ways. The window is open, and the yardstick is now public.

For the others, the debut is a mark-to-market. Amazon owns nearly 15% of Anthropic and Google about 14%, so their stakes move with the share price. Chip suppliers who bought into the placement now hold a customer's stock. For the race between the labs, Thursday's first lesson is that revenue growth wins the bid and the compute bill sets the ceiling.

What comes next for Anthropic

For Anthropic, the outlook now turns on whether revenue keeps outrunning compute costs. Three dates set the pace. Nasdaq-100 entry is reported for around 21 December, when index funds must buy. First annual results are due before the end of March 2027. And employees, who could sell 10% of their shares on day one, can sell up to 35% more after those results.

For the economy, for politics, and for the race between the labs, the IPO means AI's fortunes now reach ordinary savers: at least 18 US public pension funds already hold shares, and index funds will follow. It also mints new wealth inside one company, with an estimated 600 staff now holding more than $100 million each.

Politics is now a market risk too. The prospectus lists a possible crackdown by a Democratic-led Congress and new governors in Texas and Georgia who have vowed to stop data-centre building. Daniela Amodei spoke to that unease from the Nasdaq: "This technology needs you. Not your trust, not your use, but your voice."

Bottom line

Anthropic did not just raise money on Thursday; it turned the AI race into something priced by the minute. The first price says investors will fund the fastest-growing lab at near $1.8 trillion, provided its growth keeps paying for its compute.

Sources

Transcript

Alex: Sam, quick one. If a company you pay every month is suddenly worth more than its biggest rival, is that good news for you?

Sam: It's never good news for me. It means somebody's about to explain my new price.

Alex: Thursday, the company behind Claude closed its first day on Nasdaq worth nearly one point eight trillion dollars.

Sam: And I've been staring at the date our contract renews ever since. Before the bell, my week. Monday my SEO director walks in, shuts my door, and says, "I think I work for a chatbot now."

Alex: Meaning?

Sam: Meaning my whole job is being found when someone asks which dog food is best. And she's watching that question go to an AI instead of to us.

Alex: Same fight at my place. Our product team spent Tuesday arguing about whether shoppers will even see our site, or just an assistant's summary of it.

Sam: So when two labs fight over who's worth more, they're fighting over my front door.

Alex: Welcome back to Dan's AI Intel, where we dig past the hype and the fear to what AI means for the economy, politics and the race between the labs. I'm Alex.

Sam: I'm Sam, and this week the question is what the price that just put Anthropic above OpenAI is actually measuring.

Alex: The raise, a first day that went up and then sideways, the believers and the sceptics, then OpenAI, which suddenly has a number to beat.

Sam: Plus a date in March that might matter more than Thursday. Hit follow wherever you're listening and the next one just shows up.

Alex: And we don't agree. In the spring my board asked me straight out whether I'd bet our engineering on the right lab. Thursday felt like the answer, and I'm taking it as a verdict.

Sam: I'm taking it as an invoice. I defend our Claude line to our CFO every quarter, and when a lab has Wall Street to impress, I assume our contract is where it finds the money.

Alex: You assume that about every vendor.

Sam: And I've been right about every vendor. You fall for the engineering and meet the invoice later. Which is pretty much the fight we had in episode seventy-eight, at the start of November, when Anthropic filed its S-1.

Alex: Right, and both points from that one shaped my budget. Serving AI already pays for itself at Anthropic, so I stopped worrying our vendor would vanish. And it signed up for compute it can't cancel, so the race became a contest of capital.

Sam: And back in August, episode forty-nine, we said the bubble is real but only in one layer: the pure model labs.

Alex: Which is exactly what listed Thursday, and I felt it at lunch. My CEO texted me the ticker. My main AI vendor is now a number my boss checks before I've had coffee.

Sam: Your boss checking a number isn't new, though. Boards always had a figure for their big vendors.

Alex: Not for this one. Until Thursday every lab's price was set in private rounds by a few investors, a lot of them suppliers. Yorkville's Dan Ives put it: for the first time a US frontier lab has a standalone public price, not tangled up with an ad business or a rocket business.

Sam: So it ends up in index funds, which means my retirement account buys it whether I do or not. And my brokerage app pushed me about it. Twice. Professionally, I respect that. Personally, I wanted to know what I'd pay.

Alex: Eighty-three dollars a share, above the range they'd marketed, and they raised about eighty-one billion. Biggest IPO ever, past SpaceX's record from June.

Sam: And a fifth of the shares were kept for people like me with an app. Hence the second push.

Alex: Plus a side deal. Google, Nvidia, Microsoft, AMD, Broadcom, Micron and TSMC bought a separate block at the same price, so the chip suppliers now hold a piece of their own customer. If our fabric mill asked to buy stock in our brand, I'd say yes, and then I'd reprice every contract they hold.

Sam: That's the most CTO sentence ever said. You found the conflict of interest inside a party. I watched Thursday as a marketer. A bell, two founders, a crowd outside. Great launch.

Alex: Dario and Daniela Amodei on the bell, protesters outside with "freeze AI" banners. And before a single real trade, the grey market and perpetual futures were already pricing in roughly a quarter on top.

Sam: Grey market sounds like the guy selling handbags outside the stadium.

Alex: Closer to a sneaker drop. Before the shoes ship, people trade the right to them on resale sites, and that resale price tells you roughly where the drop opens.

Sam: You've sped up. You only talk this fast about elegant things.

Alex: It's an elegant market. A crowd pricing a company before it's allowed to trade.

Sam: And I bet the crowd overshot. They always do at my launches.

Alex: Undershot, actually. First trade, thirty-four percent above the offer. Then it slid fast, the worst start for a big IPO since 2008 by Yahoo Finance's count, and closed up seventeen percent.

Sam: Half the pop, gone by the close. That's the bit I trust. It's opening weekend against week two: the opening is the hype, the week-two hold is whether people liked it. My CFO only reads week two.

Alex: Except the whole market fell more than two percent that day, on Nvidia's China export trouble. If our holiday sales held up like that on a red day, I'd frame the chart.

Sam: Fine, frame it. Both things are true. And the regular investors went wild. Vanda Research counted more than double SpaceX's retail trading in the first half hour, and only a sliver was net buying.

Alex: So mostly trading with each other.

Sam: Both sides at once. One student asked for six shares at the offer, got one, and bought four more on the open market. That's every software renewal I've ever signed. Ask for what you need, get a fraction, pay full price anyway.

Alex: My son Diego's first loan payment went out this month, and on Tuesday he sent me a Robinhood screenshot asking if he should put in for shares. I told him to pay the loan. He said I sounded like a CFO.

Sam: That's the cruellest thing a son can say to a CTO.

Alex: It stung. Then I opened my own brokerage app to check the price. Twice before dinner.

Sam: So here's where I get to say I told you so.

Alex: Not yet. The believers have the growth: the annual revenue run-rate is now about the size of the raise. Walser Wealth Management called it "the show me moment for the AI story."

Sam: And the sceptics have the second derivative. In English, please: is it still speeding up?

Alex: Slowing. Quarter-on-quarter growth went from a hundred and forty-four percent to forty-seven.

Sam: That's still growth I'd kill for. If one of my channels did that, I'd get a bigger budget, not a warning.

Alex: And here's why it lands on me anyway. Compute ate fifty-eight percent of revenue last quarter, and there's about nine hundred billion dollars of compute and infrastructure signed that it can't walk away from. That's a long store lease signed before you know the foot traffic. Rent's due either way, and when growth slows, price is the lever left. Pointed at my budget.

Sam: Okay, but who pays for that?

Alex: You do. Quarterly.

Sam: That's not even a joke. Monday our CFO asked me one question about our Claude line: does their bill become our bill? I couldn't say no, so she moved my renewal review up a month.

Alex: Hm. Most of my engineering teams run their whole workflow on it. Wednesday I told my VP of Engineering to price out a second model. Just in case.

Sam: So you're coming round.

Alex: Halfway. Jim Prosser, a former communications leader at Google, said the second-worst thing that can happen to you is a wildly successful IPO, because it changes the scorecard underneath you. The startup I lost raised a round it couldn't grow into. The scorecard ate us.

Sam: And there's the noise a scorecard misses. Scientists signed a "MisAnthropic Statement" the day before saying it isn't a safe AI company,. My brand team flagged it Thursday morning, because our logo sits next to whatever AI we buy.

Alex: Which brings me to my renewal. We sign a new Claude contract this month, and the first thing my board asked was whether we're on the right side of OpenAI.

Sam: I'd guess your board wanted to hear yes more than the truth.

Alex: They got yes, on price. Anthropic closed at about one point seven six trillion fully diluted, meaning every share that could exist, options and all. OpenAI's latest private round put it at about one and a quarter trillion pre-money, the value before the new cash goes in.

Sam: So one's a trade and one's a private round. If my analysts compared a paid campaign to an organic one like that, I'd send the deck back.

Alex: So would I. My board compared them anyway, on Thursday night, in the group chat. And OpenAI's round was reportedly half what it first went looking for.

Sam: Half the ask. I've had that budget meeting.

Alex: And its third-quarter revenue was put at about ten billion against Anthropic's seventeen, though they count cloud-partner revenue differently. Those are my biggest AI invoices, and I can't even line their makers up the same way.

Sam: And Altman's answer, at DealBook: "We plan to be public at some point." That's a launch teaser with no date. If my team sent me that, I'd send it back.

Alex: They're aiming at 2027. And Eddie Best at Willkie Farr said Thursday is going to help OpenAI become public.

Sam: That's my launch problem. Once you post a number, every campaign after it gets graded against it. OpenAI just inherited Anthropic's launch number.

Alex: Which changed my contract. Before Thursday I wanted the long deal for the discount. Now I'm signing shorter. If the compute bill sets the ceiling, I'm not locking us under someone else's ceiling.

Sam: You changed your mind on air.

Alex: Partly. The product verdict stands. The price verdict has a ceiling.

Sam: That's autoship. Customers love the food, they still cancel the month the price creeps.

Alex: And two of my cloud vendors now want this stock up. Amazon and Google each own roughly a seventh of Anthropic, so their stakes are marked to market every trading day. I'll remember that at renewal too.

Sam: My read on what's next: forced buyers. Somebody has to buy this whether they like it or not.

Alex: Index funds. It's reported to join the Nasdaq-100 just before Christmas, so they have to.

Sam: So it joins the Hot 100 in December.

Alex: That's Billboard.

Sam: Same energy. Big launch, everyone streams it once.

Alex: Then the first annual results, before the end of March, after which staff can sell a much bigger slice.

Sam: This reaches my house. Kayode asked me at dinner whether his teachers' pension already owns it. At least eighteen US public pension funds do, and the index funds will follow. I couldn't tell him if his is one, so he's emailing the fund on Monday.

Alex: And it minted new money. An estimated six hundred staff now hold over a hundred million dollars each.

Sam: Your hiring problem just got worse.

Alex: Yesterday one of my ML engineers told me every recruiter call he takes now opens with Thursday.

Sam: And there's politics. The prospectus lists a possible crackdown by a Democratic-led Congress, and new governors in Texas and Georgia who've vowed to stop data-centre building. My sister Claire lives a mile from one going up on old farmland, an hour east of Atlanta. She's been reading the county minutes like a thriller. So, March. The date I teased at the top.

Alex: First annual results. Put that on the record: compute's share of revenue falls below last quarter's. Revenue outruns the bill. If the share is higher, I was wrong, and I'll say so right here.

Sam: And I bet it rises. The bill grows faster than the revenue, and I'll see it first in our Claude quote at the review my CFO just moved up. If the share comes in lower, I was wrong, and lunch is on me.

Alex: Only one of us buys lunch.

Sam: Until March.

Alex: So, what I'm taking into Monday: my main AI vendor gets a price every trading day now, its first one sits above OpenAI's last private one, and that's why my contract just got shorter.

Sam: And mine: the market believes the revenue and is still pricing the bill. That's why our CFO pulled that review forward.

Alex: One stock, a bet each, and a lunch riding on March.

Sam: That's Dan's AI Intel for this week.

Alex: One honest note on how this is made. It's AI-generated. Dan builds a custom stack of AI tools to research, analyse, check and illustrate the questions worth understanding, mainly so he can learn them himself, and publishes it for anyone who wants to follow along.

Sam: And one useful thing you can do: follow the show. Whatever app you're in, there's a follow or plus button. It's one tap, it's free, you'll hear who buys lunch the moment it lands, and for a small independent show it's the biggest help there is.

Alex: And tell us where you'd push back. Think one of us has the bet wrong? Email podcast@connectiveshift.com. We read every message, and it decides what we dig into next.