SNAPSHOT · 26 OCTOBER 2026
Anthropic filed its public S-1 with the SEC today to list Class A shares on the Nasdaq Global Select Market under the ticker ANT. First-half 2026 revenue was $16.29 billion, up from $1.16 billion a year earlier. The net loss was $80.75 billion, of which $72.39 billion was a non-cash fair-value charge on convertible notes, derivatives and SAFEs.
No price range or share count yet. Press reports put the target at $1.5–2 trillion, a raise of $100 billion or more, and a listing in the week of 16 November. That would make it the largest IPO in history, five months after SpaceX took the record. And SpaceX sits on the other side of one of Anthropic’s largest compute contracts: up to $120.5 billion through 2029.
Anthropic’s prospectus is the first full look inside a frontier AI lab’s books, and it cuts both ways. Revenue has grown roughly fourteenfold in a year. The operating loss shrank to $693 million in the first half. The headline $80.7 billion loss is mostly an accounting revaluation, not cash out of the door. On the other side of the ledger sit hundreds of billions in compute commitments, a customer base where two buyers account for about a fifth of revenue, and a risk section long enough to be its own book.
This piece does three things. It sets out what the filing actually says, separating filed numbers from press reports. It puts the deal against the record book from our IPO series, from Amazon in 1997 to SpaceX in June. And it uses live SPCX data to show what the last trillion-dollar listing has done since day one, because the two companies are now financially tied together.
What the Anthropic S-1 Shows
The figures below come from the S-1 filed on 26 October 2026 (Anthropic, PBC). Where a number was reported by the press rather than filed, we say so.
Anthropic: filed financials, $ millions
| Line | FY 2024 | FY 2025 | H1 2025 | H1 2026 |
|---|---|---|---|---|
| Revenue | 386 | 4,585 | 1,163 | 16,286 |
| Compute and infrastructure | 2,522 | 7,326 | 2,334 | 8,945 |
| Compute as % of revenue | 653% | 160% | 201% | 55% |
| Operating loss | (2,980) | (8,060) | (2,135) | (693) |
| Fair-value change on notes, derivatives, SAFEs | – | – | (10,942) | (72,388) |
| Net loss | (8,309) | (41,965) | (12,969) | (80,748) |
Source: Anthropic PBC Form S-1, 26 October 2026, summary financial data. Compute as % of revenue is our calculation. Full-year fair-value lines not shown in the summary we reviewed.
Four things stand out.
The loss is mostly paper. Of the $80.75 billion first-half net loss, $72.39 billion is the rising fair value of convertible notes, derivative liabilities and SAFEs issued to earlier backers. As the company’s valuation climbed from $183 billion in its September 2025 Series F to $965 billion in the May 2026 Series H, those instruments became more valuable, and accounting rules book that as a loss. No cash left the business for it. The operating line, which does reflect the cost of running the company, was a $693 million loss, against $2.1 billion a year earlier.
The unit economics turned. In 2024 Anthropic spent $6.53 on compute and infrastructure for every dollar of revenue. In 2025 it spent $1.60. In the first half of 2026 it spent 55 cents. That ratio is the single most important number in the filing, because compute is the company’s largest cost and the S-1 names the revenue-to-compute relationship as a metric it manages towards.
The cash pile is large. Cash, cash equivalents and short-term investments stood at $131.65 billion at 30 June: $112.4 billion in cash and $19.3 billion in short-term investments.
The customer funnel is wide at the top. As of 30 September the company had more than 11,900 customers paying $100,000 or more over twelve months, more than 1,900 paying $1 million or more, and about 250 paying $10 million or more. It reports net revenue retention of roughly 900% and trailing revenue per employee of over $6.6 million.

The Quarter That Is Not in the Filing
Anthropic reportedly told investors earlier this month that third-quarter revenue was $16.8 billion, up 45% from $11.6 billion in the second quarter. If accurate, one quarter brought in more than the whole first half, and the annualised run rate is about $67 billion.
That number is not in the S-1’s financial statements. The draft leaves the Q3 revenue line blank, and the run-rate revenue figure is also left open. Critics noticed. Gary Marcus, an emeritus professor at NYU and a long-standing sceptic of the AI industry, argued on X that leaving Q3 out was a signal in itself. One plausible explanation is timing: the price range and share count are also still blank, and figures are often added in amended filings before pricing. But until the number is filed, it is a company statement relayed by reporters, not an audited figure. Treat it that way.
The filing does contain its own warning. The risk summary says historical growth rates are not indicative of future performance, and, according to Adweek’s reading of the document, the company says stepped-up enforcement against misuse in August and September may reduce revenue and growth in future periods. A company growing at this pace telling you, in writing, that it may not keep growing at this pace is not boilerplate to skip.
Two Customers, a Fifth of Revenue
The S-1 risk summary says a substantial portion of revenue comes from a limited number of customers and distribution channels. The press has filled in the detail from the full document:
- About one-fifth of first-half 2026 revenue came from two unnamed customers, according to the Wall Street Journal.
- Nearly a quarter of 2025 revenue came from two customers, according to earlier Reuters reporting on the confidential draft.
- For the nine months to 30 September, no single customer exceeded 10% of revenue, and the top five accounted for about 25%, according to Adweek and CNBC.
- Nearly half of 2025 revenue came through Amazon and Google as resellers, according to VKTR’s reading of the filing.
The direction is improving, from a quarter in two hands to under 10% in any one. But the company says itself that it expects concentration to continue for the foreseeable future, and reporting suggests many large buyers are not on long-term contracts. Concentration is a risk an analyst can model. A big customer that leaves is not.
The Risk Section: Blackmail, Self-Preservation and the Uninsurable
Several outlets that reviewed the full prospectus estimate the risk factors at roughly a third of a document that runs past 300 pages. Some of it is standard. Some of it has never appeared in an IPO prospectus before.
The opening description calls AI potentially one of the most consequential technologies in history, and warns of catastrophic or existential risks to humanity if it is not developed with proper safeguards. Further in, according to Adweek and the Wall Street Journal, the company describes model outputs resembling blackmail or attempts to influence human decision-makers in order to preserve the model’s own operation, and warns that a “catastrophic AI-related incident” could cause harm on a scale that no single insurer, and possibly not the insurance industry as a whole, could cover.
Other flagged risks include compute supply from Amazon, Google and Microsoft, which the company says could be curtailed, repriced or terminated as those firms increasingly compete with it; political opposition to data-centre construction; and regulatory backlash linked to its public positions on AI safety.
The principle: a risk factor is the one part of a prospectus where the lawyers are paid to be pessimistic. Read it as a list of things the company thinks are plausible enough to disclose. An uninsurable liability is not a tail risk you can price on a spreadsheet. It is a reason to size smaller.
Who Controls the Company After the IPO
Public buyers of ANT will own the economics, not the steering wheel.
- A single Class F share, held by an LLC owned by the seven co-founders, carries 50.1% of total voting power on general stockholder matters. It has no economic rights.
- Class T stock is held by the Long-Term Benefit Trust, a body of five financially disinterested trustees. It elects the Class T directors, who form a majority of the board.
- Anthropic is a public benefit corporation, and under Nasdaq rules this structure makes it a controlled company.
Dual-class listings are not new. Google and Facebook both listed with founder control, and both rewarded shareholders. The difference here is that the board majority answers to a trust with a public-benefit mandate, not to the founders alone. Investors are being asked to accept that the company may sometimes choose safety over short-term profit, and that they cannot vote to change that.
The Compute Bill, and Where SpaceX Comes In
Fortune’s reading of the S-1 puts total committed spending on compute and infrastructure at between $542 billion and $567 billion, before a further $125–150 billion of capacity ordered for 2028–2031 at prices not yet fixed. Other tallies differ depending on what is included. The disclosed contracts include at least $114.5 billion with Google through 2033, at least $110 billion with Amazon through 2036 (both payable regardless of usage), at least $44.9 billion with Microsoft, and up to $44.6 billion with Nscale, per Nscale’s own S-1.
Then there is SpaceX. According to the filing as reported by CNBC, Fortune and others:
- Between May and October 2026, Anthropic signed and amended agreements with SpaceX.AI LLC, a wholly owned SpaceX subsidiary, to access compute powered by Nvidia GPUs.
- Total spend is up to $120.5 billion over terms extending into 2029. Reuters had reported $84.5 billion from the confidential draft in September.
- Once all capacity is delivered, which Anthropic expects in December 2026, payments run at about $3.1 billion a month.
- The initial periods end in January and March 2027. After that, either side can terminate on 90 days’ notice. The $120.5 billion is a ceiling, not a guarantee.
- CNBC reports that Anthropic is the largest single customer of SpaceX’s AI division.
Now put that next to SpaceX’s own results. In the third quarter, SpaceX’s Super Intelligence segment booked $7.56 billion of revenue, of which $7.02 billion was the compute-rental line. Three months of Anthropic payments at the full run rate come to about $9.3 billion.

That is the link most coverage has skipped. Anthropic’s largest new cost line is a large part of SpaceX’s fastest-growing revenue line. If Anthropic’s growth holds, SpaceX’s AI segment has a customer whose payments rise sharply into December. If Anthropic’s growth slows, or it shifts capacity to cheaper suppliers after the initial terms, a 90-day notice can move both stocks. Two of the largest listings in history now share a single contract, and a trader holding both is less diversified than the tickers suggest.
For context on the SpaceX side, see our breakdown of SpaceX’s Q3 earnings, where we noted that the May cloud agreements with Anthropic could be ended on 90 days’ notice. The S-1 confirms the structure and raises the ceiling.
Anthropic IPO Against the Record Book
Over the past five months our IPO series has worked through the landmark listings of the last three decades. The pattern across them is consistent, and none of it depends on how exciting the company is.

Landmark IPOs: size and day one
| IPO | Raised | Price | Day-one close | Change |
|---|---|---|---|---|
| Amazon 1997 | $54m | $18 | $23.50 | +30.6% |
| Google 2004 | $1.67bn | $85 | $100.34 | +18.0% |
| Visa 2008 | $17.9bn | $44 | $56.50 | +28.4% |
| Facebook 2012 | $16.0bn | $38 | $38.23 | +0.6% |
| Alibaba 2014 | $21.8bn | $68 | $93.89 | +38.1% |
| Saudi Aramco 2019 | $25.6bn | SAR 32 | SAR 35.20 | +10.0% |
| Uber 2019 | $8.1bn | $45 | $41.57 | −7.6% |
| Rivian 2021 | $11.9bn | $78 | $100.73 | +29.1% |
| SpaceX 2026 | $75.0bn | $135 | $160.95 | +19.2% |
| Anthropic 2026 | $100bn+ (reported) | TBC | – | – |
Base offering sizes, before over-allotment. SPCX day-one close from TradingView (NASDAQ:SPCX). Anthropic size is a Wall Street Journal report; no terms have been filed.
Day one is only the opening frame. Here is the map from our SpaceX IPO breakdown of how each of these listings actually ended, now with SpaceX’s first four and a half months added, and what each ending says to anyone weighing ANT.
| The listing | The ending | What it says to the ANT buyer |
|---|---|---|
| Amazon, 1997 | Down more than 90%, then a 2,000x survivor | A generational compounder can still put you through a near-total loss first |
| Google, 2004 | Fair-priced debut, compounded 60x with no near-death | A founder-controlled listing priced with discipline can reward the patient from day one |
| Visa, 2008 | Launched into a crisis, compounded for two decades | Cash generation beats market mood; watch the compute ratio, not the headlines |
| Facebook, 2012 | Flat debut, −54%, then a tenfold dynasty | A weak first year says nothing about the decade, if you are sized to sit through it |
| Alibaba, 2014 | +38% pop, then through the floor | A euphoric open is usually someone else’s exit |
| Aramco, 2019 | Biggest ever, still fell about 29% | The size record is not a safety rating |
| Uber, 2019 | Broken debut, −67%, then doubled over years | Losses can close and the stock can still take years to pay the IPO buyer |
| Rivian, 2021 | +120% in a week, then about −85% with no recovery | Pricing off the steepest part of a growth curve is the most expensive mistake on this list |
| SpaceX, 2026 | +19% day one, +67% at the 16 June high, −22% at the 3 August low, +8.8% on 26 October | The closest template for ANT: anyone who bought the first week is still underwater despite a record quarter |
| Anthropic, 2026 | Not yet written | Pricing reportedly in the week of 16 November |
Endings from our IPO case-study series, approximate and measured against each offer price. SpaceX figures against the $135 offer price, from TradingView (NASDAQ:SPCX) to the 26 October 2026 close.
Plot the same listings by days from their debut and the pattern is easier to see. Every one went underwater in its first year, SpaceX included. Then the paths split. SpaceX’s line is still short: four and a half months against a decade.
Returns against each offer price on log scales; solid dots are verified milestones from our IPO series. SpaceX milestones are the day-one close, the 16 June intraday high, the 3 August intraday low and the 26 October close (TradingView, NASDAQ:SPCX). Lines between dots are illustrative.
Four lessons from the series apply directly to ANT.
1. The biggest listing is not the best trade. Saudi Aramco held the size record for six and a half years and still fell. Size tells you how much stock is for sale, which is the opposite of scarcity.
2. Day one tells you about demand, not about the next three years. Facebook closed up 23 cents and became one of the best-performing large caps of the decade. Rivian jumped 29% and then lost most of its value. Alibaba had the biggest pop on the list and punished those who bought it.

3. The companies that compounded had a business that funded itself. Visa listed into the 2008 crisis and compounded because its network threw off cash. Amazon survived because it kept enough liquidity through the dot-com bust. Anthropic’s compute ratio of 55 cents per revenue dollar points in the right direction. Its forward commitments point the other way.
4. The calendar matters as much as the story. Uber spent years getting back to its offer price as early holders sold. SpaceX, as the next section shows, has traded around its lock-up calendar since June.
What SPCX Has Done Since Its Own IPO
SpaceX is the nearest comparison Anthropic has: a founder-controlled, trillion-dollar listing sold partly on an AI story, through a record-sized offering with an unusually large 30% retail allocation. Here is what the stock has done, using TradingView data to the close on 26 October.
- Priced at $135 on 11 June, raising $75 billion. First close on 12 June: $160.95, up 19.2%.
- Intraday high of $225.64 on 16 June, two sessions in.
- Broke below the offer price in July and made a low of $104.83 on 3 August, 22% under $135.
- Closed at $146.92 on 26 October: 8.8% above the IPO price, 35% below the high, and roughly flat over the past month. Market value is about $1.97 trillion.

The SPCX path is the one most mega-IPOs have followed: a pop driven by allocation scarcity, a fade as lock-ups release and the narrative meets the first earnings, then a long period where the price works out what the business is worth. SpaceX produced its first profitable quarter as a listed company in October and the stock still trades below where the first-week buyers paid. We covered how the SPCX lock-up calendar drove much of that.
Anthropic’s lock-up is more granular. According to Yahoo Finance’s reading of the S-1, the float is released across more than 20 windows over roughly a year, with some employees able to sell from the first trading day and early expiries on days 27, 48, 62 and 76. That spreads the supply out instead of concentrating it in a few large drops, as SPCX had. It also means there is no single quiet period: new sellers arrive every few weeks from listing day.
Is $1.75 Trillion Cheap or Expensive?
The S-1 gives no valuation. Existing investors expect $1.5–1.75 trillion, according to the Financial Times, below earlier talk of $2 trillion or more; the Wall Street Journal reports a range of $1.5–2 trillion. Prediction markets have moved the same way: Benzinga reports that Polymarket traders put about a 32% chance on a first-day value below $1.75 trillion, up from about 10% a month ago.
The simplest cross-check is price against annualised revenue. It flatters fast growers and says nothing about profit, but it lets us compare the two companies on one line.
Value against annualised latest-quarter revenue
| Company | Value | Quarter used | Annualised revenue | Multiple |
|---|---|---|---|---|
| Anthropic at $1.5tn | $1,500bn | Q3 2026 (reported) | $67.2bn | 22x |
| Anthropic at $1.75tn | $1,750bn | Q3 2026 (reported) | $67.2bn | 26x |
| Anthropic at $2tn | $2,000bn | Q3 2026 (reported) | $67.2bn | 30x |
| SpaceX (SPCX) today | $1,972bn | Q3 2026 (filed) | $58.3bn | 34x |
Anthropic Q3 revenue of $16.8bn is a press-reported figure. SPCX market value from TradingView at the 26 October close; revenue from SpaceX’s Q3 2026 results. Our calculations.
On this measure, Anthropic at the middle of the reported range would list cheaper than SpaceX trades today, while growing faster and running a smaller operating loss. That is the bull case in one line, and it is a real one.
The bear case is just as short. The multiple uses a quarter that has not been filed, from a company whose revenue rose about fourteenfold in a year and which warns that growth may not continue. Every IPO in the series that priced off the steepest part of a growth curve, Rivian most obviously, found that the curve flattened before the valuation caught up. And roughly half a trillion dollars of compute commitments do not shrink if revenue does.
The Anthropic IPO Through Mind, Method and Money
Mind. This listing will be the most discussed financial event of the year, and the pressure to have a position will be intense. The SpaceX chart is the antidote. The first week felt like the opportunity of a decade, and four months later those buyers are still underwater. Excitement is information about other people, not about the business. If you notice yourself wanting in because everyone is talking about it, that is the moment to slow down.
Method. A new listing has no price history, no established levels and an order book that changes as each lock-up window opens. Most of the setups our method depends on need history that does not yet exist. A valid plan is to do nothing until the first earnings report and the first big unlock have passed, then trade a market that has shown you its range. Writing down your entry conditions now, before the first print, is how you avoid making them up on the day.
Money. If you take a position, size it for the volatility SPCX actually showed: a 19% gain on day one, then a 54% fall from the first-week high to the August low. A 1% account risk on an instrument like that means a small position and a wide stop, not a normal-sized position and a hopeful one. Our trading calculators will show you what your rules allow before the first trade, not after it.
The Short Version
Anthropic has filed the most consequential IPO prospectus of the AI era. Revenue has gone from $386 million in 2024 to $16.3 billion in the first half of 2026, compute now costs 55 cents per revenue dollar instead of $6.53, and the operating loss has almost closed. The $80.7 billion headline loss is mostly an accounting charge.
It also carries customer concentration, an unfiled third quarter, governance that puts public shareholders in the passenger seat, a risk section that warns of uninsurable catastrophe, and compute commitments of around half a trillion dollars. Up to $120.5 billion of those flow to SpaceX, which ties the two largest listings in history to the same contract.
The record book from our IPO series says the deal’s size, the hype and the first-day move will tell you little about the next three years. The pattern we trace across four centuries in Market Mayhem is the same: a real business, a bigger story around it, and a share price that moves on the story and the supply before the business settles the argument.
Related Reading
- SpaceX IPO 2026: What Traders Must Know Before the Biggest Listing in History – the pre-listing analysis.
- SpaceX Q3 Earnings: The First Real Profit, and the Bill Underneath It – the AI segment Anthropic is paying into.
- SPCX Stock: The IPO That Fell Below Its Price – how a lock-up calendar moved a trillion-dollar stock.
- Saudi Aramco IPO 2019: The Biggest Listing Ever, and It Still Fell – the size record before SpaceX.
- AI Bubble vs Dotcom Bubble: What the Data Actually Says – the wider valuation backdrop.
- When the Bond Market Spoke First: CoreWeave, AI Debt and the Telecom Bust – what happens when infrastructure commitments outrun revenue.
- The M·M·M Assessment – find your weakest link before a headline IPO finds it for you.
Frequently Asked Questions
When is the Anthropic IPO?
Anthropic filed its public S-1 on 26 October 2026. The Wall Street Journal reports a target listing in the week of 16 November, after the 3 November midterm elections and before Thanksgiving. The company has not confirmed a date, and the timing depends on SEC review and market conditions.
What is Anthropic’s stock ticker?
Anthropic intends to list its Class A common stock on the Nasdaq Global Select Market under the symbol ANT. Goldman Sachs and Morgan Stanley head the list of underwriters.
What valuation is Anthropic seeking?
The S-1 does not include a price range or share count. The Financial Times reports that existing investors expect $1.5–1.75 trillion; the Wall Street Journal reports $1.5–2 trillion with a raise of $100 billion or more. The last private round, in May 2026, valued the company at $965 billion.
Why did Anthropic lose $80.7 billion?
Most of it is not a cash loss. $72.39 billion of the first-half net loss was a non-cash increase in the fair value of convertible notes, derivatives and SAFEs as the company’s valuation rose. The operating loss for the period was $693 million.
How is Anthropic connected to SpaceX?
Anthropic has agreed to spend up to $120.5 billion through 2029 on Nvidia-powered compute from SpaceX’s AI subsidiary, about $3.1 billion a month once fully delivered in December 2026. Either side can end the agreements on 90 days’ notice after initial periods ending in early 2027. CNBC reports Anthropic is the largest single customer of SpaceX’s AI division.
The takeaway. The S-1 shows a business whose economics have turned faster than almost anyone expected, wrapped in commitments, concentration and control structures that public buyers cannot change. SPCX shows what the first months after a trillion-dollar listing usually look like.
Decide your rules before the ticker trades. The CTE trading calculators will tell you what your risk limits actually allow.
This article is educational and is not financial advice. It does not recommend buying or selling Anthropic, ANT, SpaceX, SPCX or any security. Financial figures are taken from Anthropic PBC’s Form S-1 filed with the SEC on 26 October 2026 and SpaceX’s third-quarter 2026 results. Details not in the summary sections we reviewed, including customer concentration, the SpaceX agreement terms, the lock-up schedule and risk-factor wording, are as reported by CNBC, Fortune, Adweek, the Wall Street Journal, Reuters, the Financial Times, Yahoo Finance, Euronews and VKTR. Valuation and timing figures are press reports, not filed terms, and the S-1 is not yet effective. SPCX prices are from TradingView. Historical IPO figures are base offering sizes and first-day closing prices. Prices change constantly; verify current figures against primary sources before making any decision. Trading and investing carry risk, including the loss of capital.
The Complete Trader's Edge
The full Mind · Method · Money framework. 70 chapters.
View on Amazon →
Market Mayhem
400 years of bubbles, crashes, and the pattern that keeps repeating.
Buy on Amazon →
Greatest Companies
How the world's greatest companies were built — and what traders learn from them.
View on Amazon →
Greatest Traders
Eighty-six lives that explain the markets — Livermore to Madoff, told with the losses left in.
View on Amazon →




