Anthropic, the company behind the Claude family of AI models, is heading for the public markets — and the week ahead may decide how much the most anticipated AI listing of 2026 is actually worth. The company filed its original Form S-1 with the US Securities and Exchange Commission on October 26, 2026, followed by an amended S-1/A on November 2 that added third-quarter results. It plans to list Class A shares on the Nasdaq Global Select Market under the ticker ANT, with Goldman Sachs and Morgan Stanley leading a 22-bank underwriting syndicate. The formal investor roadshow is expected to begin the week of November 9, with a first trade targeted before Thanksgiving on November 26 — but after a brutal week for technology stocks, the valuation investors are willing to pay has fallen fast, and even the timetable is now in doubt.
What the amended S-1 actually says
The S-1/A, filed November 2 (accession 0001628280-26-068676), is the first public look at Anthropic’s third quarter, and the numbers are extraordinary in both directions. Revenue for the three months ended September 30 was $16.96 billion, up 1,127 percent from the same quarter a year earlier and up about 47 percent from the $11.6 billion posted in the second quarter. For the first nine months of 2026, total revenue reached $33.24 billion, roughly twelve times the comparable period in 2025. No AI company has previously reached this scale of revenue this quickly.
But the growth rate is also the first concrete sign of deceleration. Sequential growth had accelerated for three straight quarters — 48 percent in the fourth quarter of last year, 132 percent in the first quarter, 144 percent in the second — before dropping to about 47 percent in the third. The filing attributes the slowdown to capacity constraints while demand outpaced available compute, seasonal softness in summer business activity, and restrictions on customers found to have violated Anthropic’s terms of service. It also discloses that price reductions on new model versions and contractual discounts cut third-quarter revenue by roughly $600 million.
The operating picture flipped as well. Anthropic posted a $567 million operating loss in the third quarter, reversing the $974 million operating profit of the second quarter. The company points to two items: about $5 billion of compute bought on expensive short-term contracts, and a $2 billion non-cash charge tied to its employee donation-matching program. Adjusted operating income, which strips out those items, was $2.5 billion — a 15 percent margin, down from 19 percent in the second quarter. On a reported GAAP basis the quarterly net loss was $23.29 billion, but almost all of it reflects non-cash write-downs tied to the rising fair value of convertible notes from earlier fundraising rounds — an accounting effect that largely disappears upon the IPO. Operating cash flow was positive at $6.6 billion, and the company held $146.6 billion in total liquidity at September 30.
The $895 billion compute commitment
The single most striking number in the filing is not the revenue or the loss. It is the $895.4 billion in total purchase and lease commitments for compute and infrastructure, with roughly $447.5 billion expected to be incurred before the end of 2030 and about $69.8 billion due within 12 months. The obligations extend through 2045 and are substantially all for compute and data center infrastructure.
The scale of the step-up is worth pausing on. The original S-1 filed October 26 disclosed approximately $265.6 billion in compute commitments as of June 30. In roughly the same span that quarterly revenue grew by $5.4 billion, contractual compute obligations grew by more than $600 billion. One nuance matters: analysts at Cape Fear Advisors, reading the S-1/A directly, identified $213.8 billion of the total as conditional or terminable rather than fixed-pay obligations. Even so, the confirmed portion is an infrastructure commitment at a scale no technology company has previously made.
The strategy behind it is named in the filing as Anthropic Direct — an asset-light model in which partners secure the land, power and facilities while Anthropic leases the computing capacity inside. The goal is to replace expensive short-term compute with cheaper dedicated infrastructure as new facilities come online. Prior agreements include arrangements with SpaceX, Amazon Web Services, Google, Microsoft, Broadcom and Nscale, with SpaceX providing up to $120.5 billion in Nvidia GPU compute through 2029. Compute and infrastructure costs already reached about $10 billion in the third quarter alone — 58 percent of revenue, up from 48 percent in the second quarter. That is the core tension of this IPO: the more Claude is used, the more it costs to run, and the business has to prove that locking in dedicated capacity at scale will compress per-token costs faster than usage grows.
Why the valuation talk fell from $2 trillion
Anthropic has not set an official IPO valuation — the price range and share count in the November 2 S-1/A are still blank, and no date has been filed. The last confirmed valuation is $965 billion post-money, from the $65 billion Series H announced on May 28, 2026. Everything else is reported talk from people familiar with investor discussions, and that talk has fallen with every report since early October, when figures approaching $2 trillion were in circulation.
Reuters and the Wall Street Journal reported that prospective buyers had settled on $1.5 trillion to $1.6 trillion after the amended filing landed. Then came the market. On Thursday, November 5, the Nasdaq 100 fell 6.4 percent — its biggest one-day drop on record — erasing about $2.3 trillion of market value, two days after midterm elections in which Democrats won both chambers of Congress and data-center opposition became a campaign issue in several states. Bloomberg reported on November 6 that prospective buyers were now discussing $1.2 trillion to $1.3 trillion, with some as low as $1.1 trillion. The Wall Street Journal separately reported three prospective investors saying the price likely has to come down, with one near $1 trillion. Anthropic declined to comment to Bloomberg, Reuters and CNBC.
The filing itself warns that opposition to data-center development “has also become more organized and coordinated across jurisdictions,” naming measures in New York and Texas. For a company with $895 billion of compute commitments, slower permitting is a direct risk to the growth story the valuation depends on.
The roadshow and the road to Nasdaq
The reported sequence is tight. A roadshow could start the week of November 9, with pricing the following week and a first trade as early as the week of November 16, targeted for the week before Thanksgiving. But Reuters reports that people familiar with the investor talks question whether the roadshow can start on time after the selloff, and prediction markets have moved: Kalshi’s odds of an IPO announcement before December fell to 57 percent from more than 85 percent in a week, according to CNBC. As of November 7, Anthropic had not withdrawn its registration statement — a company abandoning an offering normally files a Form RW, and there is none — and had not announced a delay. The risk is a delay, not a cancellation, and companies often wait out a volatile market and update the same registration statement later.
One unusual feature of the deal deserves attention from anyone watching the first days of trading. Under the proposed lock-up arrangements, certain non-founder directors, officers, current and former employees and consultants may transfer up to 10 percent of their locked securities from the first day of trading, with an additional 35 percent becoming transferable after the company’s full-year 2026 results are released. That does not mean 10 percent of the entire share capital can be sold immediately, and permission to sell is not a promise to sell — but alongside a limited public float, early insider selling is a variable that could sharpen price moves in either direction.
There is also a governance point retail investors should understand before they ever see a share price: the Class A shares being sold carry economics, not control. The Long-Term Benefit Trust and the founders’ single Class F share control the board and general votes. And most retail buyers will not get the IPO price at all — allocations go mainly to institutions, so most people will buy ANT after it opens on Nasdaq, often well above the offer price on hot deals. Ignore any “pre-IPO Anthropic shares” offers on social media or messaging apps; they are a common scam pattern.
What investors should watch next
Five things will decide how this listing lands. First, an S-1/A with a price range — multiply the midpoint by the post-offering share count and you get the first official valuation. Second, the offering size: Reuters says some expected up to $100 billion, while Bloomberg says Anthropic may start with a conservative size and valuation. Third, the customer base: the top five customers accounted for about 25 percent of revenue in the first nine months of 2026, no single customer exceeded 10 percent, and about 44 percent of nine-month revenue came through cloud provider marketplaces — and many large enterprise users lack multi-year lock-in terms, so they can scale back usage. The counterweight is retention: net revenue retention is approximately 900 percent, meaning existing customers are spending roughly nine times more than a year ago, and 79 percent of third-quarter revenue came from outside the top five, up from 64 percent in the first quarter. Fourth, whether the Nasdaq stabilises after the selloff — a stable market makes a November pricing more likely. Fifth, fresh revenue data: Reuters says investors are waiting for an October revenue figure before committing.
The valuation math is the other way to frame the debate. Price-to-sales is the only workable yardstick, because GAAP earnings are distorted by non-cash charges. At $1.3 trillion, buyers would pay about 19 times the current quarter annualised; at $2 trillion, about 29.5 times. The multiple compresses only if growth continues — if revenue keeps growing about 47 percent a quarter, today’s price looks cheap within a year; if growth decays the way the third quarter hints, the multiple stays high for longer. And the roughly $69.8 billion of commitments due in the next 12 months exceeds even annualised third-quarter revenue, so margins after compute will decide what the multiple is really worth.
What it means for the AI industry
Anthropic is not listing in a vacuum. The same week brought a crowded field of AI milestones: Mistral released the weights of its 1-trillion-parameter Large 4 model for free download, Google made its Gemini Omni 1.2 Flash video model generally available with 15-second clips at a 30 percent price cut, OpenAI agreed to buy Pierre to give its Codex coding agent a Git home, and NVIDIA shipped Cosmos 3 Nano world models so robots and cars can practice in simulation. The AI race is now being fought on infrastructure, distribution and cost — not just model quality — and Anthropic’s filing is the clearest public window yet into what that fight costs.
At a reported $1.5 trillion, this would be one of the largest IPOs in US market history, though well below the $86.2 billion SpaceX raised in its June debut at an approximately $1.77 trillion valuation. The deeper test is whether public markets will fund AI infrastructure at this scale on the terms Anthropic is asking. If the roadshow prices near the top of the reported range, it validates the bet that demand for AI will keep compounding. If it slips to December or prices far lower, it will be read as the market putting a ceiling on the AI buildout — and every supplier, from the cloud providers to the chipmakers, will feel it.
Bottom line
Anthropic’s IPO has gone from a question of how high to a question of how much lower, and when, in the space of a week. The filing shows a business growing at a pace no AI company has reached before — $17 billion in a single quarter, up more than elevenfold year over year — carrying a compute bill no company has ever signed, into a market that just had its worst day on record. The roadshow starting the week of November 9 is where that tension gets priced. Watch for the price range in the next S-1/A: it will be the first official number in a story that has so far been told entirely in reported ranges and blank spaces.
Related reading and sources
- Mistral Large 4 goes open weight: Europe’s 1-trillion-parameter model is free to download
- Google Gemini Omni 1.2 Flash: 15-second AI video at a 30 percent price cut
- OpenAI buys Pierre: Codex is about to get its own Git home for AI-written code
- NVIDIA Cosmos 3 Nano: four open world models that let robots and cars practice in simulation
- TechTimes: Anthropic IPO filing shows revenue growth slowing as investors cut target valuation
- Decode the Future: Anthropic IPO valuation and date tracker
- GO Markets: Anthropic’s IPO reality check — the $895 billion compute bill and the risks traders are watching
