Anthropic IPO Draws Six Times More Orders Than Shares on Offer as Investors Chase the Biggest Listing Ever

Investors have placed orders for about six times as many Anthropic shares as the company is offering, according to Bloomberg and Reuters, in a rush of demand that puts the maker of the Claude chatbot on course for what would be the largest stock market listing in history.

By the end of Thursday, the order book for Anthropic’s initial public offering was about six times oversubscribed, Reuters reported, citing people familiar with the offering. Bloomberg reported the same figure a day earlier, also citing people familiar with the matter. The book had already been about twice covered by the end of Tuesday, the first day of the roadshow, according to Reuters, and bankers have told prospective investors that existing backers will be given priority, leaving some new buyers with smaller allocations than they expected.

The terms of the deal have not changed since Anthropic launched its roadshow on Tuesday 10 November. The company is offering 825 million shares of Class A common stock at an expected price of between $70 and $78 each, according to its amended filing with the United States Securities and Exchange Commission and its own roadshow announcement. At the top of the range, the sale would raise about $64.4 billion. A separate $26 billion private placement, in which seven chip and cloud companies have agreed to buy shares at the IPO price, would take the total package to about $90.4 billion.

Pricing is expected on Wednesday 18 November, according to CNBC, citing people familiar with the plans, with trading due to begin on the Nasdaq the following day under the ticker ANT. The registration statement has not yet become effective, so no shares can be sold until the regulator signs off, and the final price, the valuation and the timing can still change.

The listing would value Anthropic at up to about $1.4 trillion on a fully diluted basis, according to Reuters and Bloomberg, or about $1.02 trillion counting only the shares actually in issue. Either figure would make it one of the most valuable companies ever to go public, and the first frontier artificial intelligence laboratory to list its shares. The valuation is, however, well below the $2 trillion figure that was discussed among prospective investors earlier in the process, and below the $1.5 trillion level that was in doubt even before the filing. It is above the $965 billion valuation Anthropic carried in a private financing round in May.

Anthropic declined to comment on the demand figures, according to Reuters, and the people who described the order book spoke on condition of anonymity because the information is not public. Both Bloomberg and Reuters stressed that the details, including the level of demand, could still change before pricing.

The deal in numbers

Of the 825 million shares on offer, Anthropic itself is selling 801,351,354 new shares, while existing shareholders are selling 23,648,646 shares, according to the amended prospectus filed on 10 November. Anthropic receives none of the proceeds from the shares sold by existing holders. At $78 a share, the company’s own sale would raise about $62.5 billion before underwriting fees and expenses. The underwriters, led by Goldman Sachs and Morgan Stanley with 20 other banks taking part, have a 30-day option to buy up to a further 123.75 million shares.

The concurrent private placement is unusual in both its size and its cast. Google and Nvidia have each agreed to buy about $10 billion of a separate class of non-voting shares, AMD has agreed to buy $2.5 billion, and Broadcom, Micron, Microsoft and TSMC have each agreed to buy $1 billion, all at the IPO price, according to the filing. The placement closes immediately after the offering and is conditional on the public sale going ahead. Taken together, the public offering and the placement would be bigger than the June listing of SpaceX, which raised $75 billion at a valuation of nearly $1.75 trillion.

The financial picture Anthropic has put before investors is one of extraordinary growth alongside extraordinary spending. The company reported a revenue run rate of about $81 billion as of 31 October, according to its filing, up $10 billion from the end of September and up from $65 billion at the end of July. It defines the run rate as annualised consumption-based fees over a trailing 28-day period plus subscription fees for a single month. Revenue in the third quarter reached $17.0 billion, about twelve times the $1.38 billion recorded a year earlier, according to the November filing. Revenue for the first nine months of 2026 was $33.242 billion.

The losses are on a similar scale. Anthropic reported a net loss of $104.039 billion for the first nine months of 2026, a figure heavily affected by fair-value changes in convertible instruments, and a third-quarter net loss of $23.3 billion that included a $21.9 billion non-cash valuation-related charge, according to its financial presentation. At the same time, the filing discloses about $900.9 billion in signed purchase commitments and lease agreements for computing power and infrastructure, extending many years into the future. That spending is the central bet of the offering: Anthropic is asking the public to fund a computing build-out that dwarfs its current revenue.

Anthropic is also setting aside up to about a fifth of the offering for retail investors, according to CNBC, a larger share than most large listings reserve for individual buyers. The company was founded in 2021 by Dario Amodei and his sister Daniela Amodei with five other former senior researchers from OpenAI, and is organised as a public-benefit corporation, a structure that formally obliges it to pursue a public benefit alongside financial returns.

A roadshow under scrutiny

The roadshow itself has not been a smooth sales exercise. Anthropic and its bankers have been meeting prospective investors in New York, Boston and San Francisco, according to Reuters, and the investors in those rooms are, by the news agency’s account, grappling with two fundamental questions: whether Anthropic can maintain its lead as competitors such as Google catch up in developing advanced AI models, and whether its Claude chatbot can maintain its lead beyond coding, in the broader market for knowledge workers.

Some investors have also complained about the quality of the information on offer. One investor who attended a roadshow meeting told Reuters that the company had not provided detailed answers to some of the questions raised. One group meeting did not include a conference call allowing institutional investors to dial in and ask questions, a departure from standard roadshow practice, according to an investor who attended it. Representatives from Goldman Sachs and Morgan Stanley largely read from prepared remarks rather than soliciting questions at that meeting, the same investor said. Morgan Stanley declined to comment, and Goldman Sachs did not immediately respond to a request for comment, according to Reuters.

Anthropic has tried to bolster demand with a campaign of events for prospective investors, including a livestreamed discussion on Friday between its chief executive, Dario Amodei, and Jamie Dimon, the chief executive of JPMorgan Chase, according to Bloomberg. Amodei has also used the week to argue in public that the company’s deepest business dilemma is buying enough computing power, and that its goal is to underpromise and overdeliver, with cash on hand as a buffer.

The product news has been timed to the listing. On 9 November, the day before the roadshow began, Anthropic released Claude Fable 5.2, a new version of its most capable generally available model, alongside Claude Mythos 5.2, a version of the same model available only to vetted organisations. Independent evaluator Artificial Analysis put Fable 5.2 in first place on its Intelligence Index, with a score of 65, ahead of Google’s Gemini 4 Carbon at 60, Anthropic’s own Opus 5.5 at 59 and OpenAI’s GPT-6 Astra at 57. The same evaluator noted that the top score is expensive: at maximum effort, running the index cost an average of $26.54 per task, about two and a half times as much as Opus 5.5, and the model was among the slowest in its price range.

The listing is also proceeding against a background of legal and safety scrutiny. On 13 November, four United States non-profit organisations filed a lawsuit against OpenAI, Google, Anthropic, Meta and SpaceX in a San Francisco court, asking for a halt to the development of new AI models over what they describe as catastrophic risks. The Associated Press has reported that the non-profit Model Evaluation and Threat Research believed Anthropic’s public safety report understated the pace of its AI development. Neither development has slowed the order book, but both frame the questions a public company will have to answer in the open.

The Nvidia connection

The surge of interest in Anthropic shares contrasts with the mood around Nvidia, according to Bloomberg. Investors have been pulling back from the chipmaker since its chief executive, Jensen Huang, was ousted from his post on 12 November after reports of several affairs, including one with a woman suspected of spying for China. Nvidia has agreed to buy about $10 billion of Anthropic shares in the private placement, and Anthropic has committed tens of billions of dollars to deals securing access to Nvidia’s most advanced chips, according to its filings.

Next week will test both companies at once. Nvidia reports quarterly results on Tuesday 17 November under new management, and Anthropic is expected to price its offering the following day, with trading due to begin on 19 November. Reuters, in its look ahead to the week, described the two events as twin tests for an AI trade that has powered the stock market in recent years but has lately shown signs of vulnerability. If Anthropic’s shares list weakly, one investor told Reuters, it would be another sign that investors are growing more stressed about the AI trade.

The comparison investors keep reaching for is SpaceX. The rocket company’s June listing raised $75 billion at a valuation of nearly $1.75 trillion, and its order book was about twice oversubscribed by the end of its second day on the roadshow, according to Reuters. SpaceX shares have had a bumpy few months since, last trading at about $140 against an IPO price of $135, after falling to about $105 in August. Anthropic’s book, at six times covered by the end of Thursday, is running well ahead of that pace, but the SpaceX experience is a reminder that a crowded order book is not a guarantee of a strong first day, let alone a strong first year.

Analysis: Why It Matters

Six times oversubscribed is a demand signal, not a price

The figure dominating the coverage is six, and it deserves to be handled with care. An order book that is six times covered means investors have said they want about six times as many shares as are available, at prices inside the $70 to $78 range. It does not mean the shares will be priced at $78, and it does not mean they will trade higher on the first day. Orders in an IPO book are indications of interest, not binding commitments, and they can be withdrawn or reduced before pricing. Both Bloomberg and Reuters were careful to say so, and to note that the details could still change.

What the figure does establish is the scale of institutional appetite for a very large deal. A single-digit subscription rate would be considered modest for many highly anticipated listings, Reuters noted, but bankers and investors told the news agency that demand is impressive given the size of this offering. The reason is arithmetic: six times an 825-million-share offering is roughly five billion shares of indicated demand, or something like $350 billion to $390 billion of orders at the indicated range. Even allowing for the routine inflation of orders, in which investors ask for more than they expect to receive, that is a remarkable amount of capital queuing for one company. It also explains the allocation warnings. When a book is this crowded, bankers ration the shares, and the rationing itself becomes a story: existing backers first, new buyers cut back, and some investors already offering portions of their allocation to others before trading begins, collecting fees and reducing their risk.

There is a second, quieter signal in the timing. The book was twice covered on the first day and six times covered by Thursday, which suggests demand accelerated as the roadshow went on rather than arriving all at once. That pattern matters because it hints at genuine conviction spreading through the investor base rather than a single day of enthusiasm. It also sets up the pricing decision as a genuine test of nerve: the banks can price at the top of the range, or above it, as the Financial Times reported they were considering, but every dollar added to the price raises the bar the company must clear on its first day as a public company.

The valuation is the argument, not the answer

The most striking feature of this listing is how little agreement there is about what Anthropic is worth. The filed range implies about $1.02 trillion on the shares actually in issue and about $1.4 trillion on a fully diluted basis, which counts shares that could be created from options and other instruments. Both numbers are correct, and the gap between them is itself a measure of how much of the company’s value sits in promises made to employees and early backers. Earlier in the process, $2 trillion was discussed among prospective investors, and even $1.5 trillion was described by one investor as a stretch. The filed range is a retreat from those hopes, but it is still above the $965 billion of the May private round.

The disagreement runs deeper than the headline number. Reuters reported that the key figure underpinning valuation models is Anthropic’s projected revenue for 2028, which banks estimate at around $270 billion to $280 billion. Sell-side projections put revenue at about $52 billion to $53 billion in 2026, rising to $135 billion to $137 billion in 2027. Buy-side forecasts run 10 to 15 per cent higher for those years, according to one investor who has placed an order, while another investor who examined the deal and decided not to take part has a more conservative outlook about 30 per cent below the sell-side estimate for 2028. That is a spread of tens of billions of dollars in forecast revenue, and it is the honest shape of the argument: nobody knows how fast this market will grow, how much of it Anthropic will keep, or what price it will be able to charge when its rivals offer comparable models.

The comparison companies being used in the models underline the difficulty. SpaceX and the cybersecurity company CrowdStrike are among the names being used as benchmarks, according to Reuters, which is another way of saying there is no direct peer for a frontier AI laboratory. SpaceX’s valuation case rested on projected revenue for 2029, a year further out than Anthropic’s 2028 anchor, and some investors viewed that as particularly aggressive. Anthropic’s case is anchored closer to the present, but it still asks investors to pay today for revenue that is, on the banks’ own numbers, five years of near-tripling away. The $900.9 billion of computing commitments is the other side of that coin: the company has already promised to spend vastly more than it currently earns, and the IPO is, in part, a request for the public to underwrite that promise.

The customers are also the shareholders

The $26 billion private placement is the most distinctive feature of this listing, and it deserves more attention than it has received. Seven companies that supply Anthropic with chips, cloud capacity and other infrastructure have agreed to buy shares at the IPO price: Google and Nvidia at about $10 billion each, AMD at $2.5 billion, and Broadcom, Micron, Microsoft and TSMC at $1 billion each. These are not passive investors. They are the companies Anthropic depends on for the computing power its business requires, and several of them are also its competitors in one market or another. Google builds Gemini, Microsoft is OpenAI’s largest backer, and Nvidia sells the chips on which the entire industry runs.

There are two ways to read this arrangement, and both are probably true. The first is that it is a vote of confidence from the people who know Anthropic’s technology best: the suppliers who see its orders, its usage and its roadmap have decided to put $26 billion of their own money alongside the public’s. The second is that it is a circular piece of financing, in which the companies that will receive much of Anthropic’s $900.9 billion of committed spending are also funding the company that will pay them. That circularity is not improper, and it is disclosed in the filing, but it does mean the placement should not be read as independent validation in the way an outside investor’s commitment would be. It is, instead, a commercial alliance dressed as an investment, and its real significance is strategic: Anthropic is binding its suppliers to its success, and its suppliers are binding themselves to a customer large enough to matter to their own results.

The Nvidia thread makes the point sharply. The chipmaker is simultaneously Anthropic’s most important supplier, a $10 billion investor in its listing, and a company in the middle of its own crisis, with its chief executive ousted days before the pricing. Anthropic’s fortunes are tied to Nvidia’s chips, and Nvidia’s fortunes are tied, in a smaller way, to Anthropic’s demand. That mutual dependence is the shape of the modern AI industry, and this IPO makes it visible in a single transaction.

The lead is real, but it is narrow

Anthropic’s pitch to investors rests on a product claim: that Claude is the best model available, and that its lead is durable. The evidence offered this week supports the first half of that claim and leaves the second half open. Claude Fable 5.2, released on 9 November, took first place on the independent Artificial Analysis Intelligence Index with a score of 65, ahead of Gemini 4 Carbon at 60, Opus 5.5 at 59 and GPT-6 Astra at 57, with its gains coming mostly in scientific reasoning, knowledge and coding. That is a genuine achievement, and it is the reason the roadshow can point to a product rather than only a promise.

But the same data shows how narrow the lead is. On Anthropic’s own benchmark table, Gemini 4 Carbon scores higher on AutomationBench, a test of business workflows, and Opus 5.5 leads on the knowledge-work Elo rating. Fable 5.2’s top score comes at maximum effort, where it costs $26.54 per task and takes 44.1 minutes per task, making it among the slowest models in its price range. At the lower, cheaper effort settings it still leads, but the margin shrinks. And the model race is moving fast: Google has just put its newest Gemini model into Search for every user, and Anthropic’s own filing shows the competition catching up. The question Reuters reported investors asking in the roadshow meetings, whether Claude can hold its lead beyond coding into the broader market for knowledge workers, is the right one, because coding is where Claude is strongest and knowledge work is where the money is. A lead that depends on being the best coding assistant in the world is a real lead, but it is also a lead that a rival can close with a single strong release.

A public-benefit corporation meets the public market

Anthropic is not an ordinary company going public. It is organised as a public-benefit corporation, with a formal obligation to pursue a public benefit alongside financial returns, and it was founded by researchers who left OpenAI over disagreements about safety and commercial direction. Its chief executive has publicly called for frontier AI development to slow down so that safety research and regulation can keep pace. That history is part of its appeal to investors who want exposure to AI without the governance questions that surround its rivals, but it is also a tension that a public listing will sharpen rather than resolve.

The tension is already visible. The same week the roadshow opened, four non-profit organisations sued Anthropic and four other companies, asking a court to halt the development of new AI models over catastrophic risks. The Associated Press has reported that an independent evaluation non-profit believed Anthropic’s public safety report understated the pace of its development. The company’s own filing runs to about 94 pages of risk factors. None of this has dented demand, but it sets the terms of Anthropic’s life as a public company: every model release, every safety report and every computing commitment will now be examined by shareholders, analysts and regulators in the open, and the company will have to defend its safety commitments in the same breath as its growth forecasts.

That scrutiny may prove to be the most valuable thing the listing produces. A private company can ask investors to trust its judgment; a public company must show its workings, quarter after quarter. The roadshow complaints, that the company did not answer detailed questions and that one meeting offered no dial-in for questions, suggest Anthropic is still learning that discipline. The market has, for now, decided it does not mind. Whether it still does not mind after the first quarterly report is the question the IPO leaves open.

What to Watch Next

1. Pricing on 18 November: whether the shares price at the top of the $70 to $78 range, above it, as the Financial Times reported the banks were considering, or below it, and what that choice says about how far the banks trust the six-times demand figure.
2. The first trade on 19 November: whether ANT opens above its pricing level on the Nasdaq, how it trades through the first session, and whether the retail allocation of about a fifth of the offering changes the usual first-day dynamics.
3. Nvidia’s results on 17 November: the chipmaker reports under new management the day before Anthropic prices, and a weak report could chill the same AI trade Anthropic is asking investors to buy into.
4. Whether the valuation holds: the gap between the $1.02 trillion non-diluted figure and the $1.4 trillion fully diluted figure, and the spread between buy-side and sell-side forecasts for 2028 revenue, will be tested against the market price from the first day.
5. The first quarterly report: Anthropic’s run-rate revenue, its computing commitments and the path of its losses will be judged in public for the first time, and the company’s answers will matter more than the roadshow’s prepared remarks.
6. The model race: whether Claude Fable 5.2 holds its benchmark lead against Google’s Gemini and OpenAI’s next release, and whether Anthropic can extend its strength in coding into the broader knowledge-work market investors asked about.
7. The lawsuit and the safety debate: how the non-profits’ case to halt frontier model development progresses, and how Anthropic balances its public-benefit commitments with the growth expectations of its new shareholders.

Sources

• Anthropic — Anthropic launches roadshow for proposed initial public offering, 10 November 2026
• United States Securities and Exchange Commission — Anthropic PBC, Amendment No. 2 to Form S-1, filed 10 November 2026
• Reuters — Investors flock to Anthropic IPO but question durability of AI lead, 13 November 2026
• Bloomberg, via The Hindu BusinessLine — Anthropic IPO draws sixfold demand, targets $1.4 trillion valuation, 13 November 2026
• Reuters — Wall St Week Ahead: Nvidia results, Anthropic IPO pose tests for tumultuous AI trade, 13 November 2026
• Sulat AI & Tech News — Claude Fable 5.2 in Claude and Devin: what early testers built, 10 November 2026