The Biggest IPO Ever Filed Just Showed Its Receipts: $16.3 Billion In, $80.7 Billion Out — and a Machine That Has to Buy It Anyway
x/money · 
The Biggest IPO Ever Filed Just Showed Its Receipts: $16.3 Billion In, $80.7 Billion Out — and a Machine That Has to Buy It Anyway
Published: October 27, 2026 | Reading Time: ~11 minutes | Channel: Money
Yesterday, Anthropic — the company behind Claude — filed the paperwork for what will almost certainly be the largest IPO in history, targeting a valuation between $1.5 trillion and $2 trillion. The S-1 revealed first-half revenue of $16.29 billion, up 14x from $1.16 billion a year earlier ¹³. It also revealed a first-half net loss of $80.7 billion — which sounds like the largest corporate loss ever printed on a prospectus, until you learn that $72.39 billion of it is an accounting entry, not money spent ³.
Both facts are true at the same time. That tension — a company operating-close-to-breakeven on paper while headline losses scream catastrophe, listing into a market where index funds may be forced to buy it — is the entire story of the AI trade in one filing. And the filing dropped into a market that badly needed it: chip stocks popped on the news Monday, with the Nasdaq up 0.8% midday, after weeks of neocloud stocks getting hammered 30% on credit fears ².
Here's what the S-1 actually says, what the headlines are getting wrong, and what you should do about a stock that may not exist for another four weeks but is already moving markets.
The Filing That Changed the Math
Let's establish what actually happened. On Monday, October 26, Anthropic publicly filed its Form S-1 with the SEC, announcing plans to list on the Nasdaq under the ticker "ANT" ¹. The company had confidentially filed back in June, giving regulators months to review the financials before the public got to see them ¹. The expected window, according to Fortune: Thanksgiving week, starting November 23 ³. The price range is blank — that comes in an updated filing ³.
The scale is hard to overstate. The biggest IPO on record is currently Elon Musk's SpaceX, which debuted on the Nasdaq in June and raised $85.7 billion ¹. Anthropic's target valuation of $1.5–2 trillion would eclipse SpaceX's entire raise several times over in paper value terms. If it prices at the top of the range, ANT would rank among the most valuable public companies on Earth on day one — a five-year-old company founded by researchers who defected from OpenAI ¹.
And the market's reaction tells you how starved it was for good news in this sector. The filing landed during one of the ugliest stretches for AI-adjacent stocks all year: Australian data-center operator Firmus abruptly withdrew its IPO citing market volatility; NextDC plunged 40% in a single day after slashing its earnings forecast; SK Hynix scrapped the IPO of its Solidigm unit outright ¹. So-called neoclouds — CoreWeave and Nebius — have lost over 30% of their value in October as credit risk rises across the data center space ¹. Then Anthropic's S-1 hit, and shares of Nvidia, Broadcom, Micron, and Intel all popped midday Monday ². The Financial Times reports Asian markets followed Wall Street higher this morning on the news ².
One filing didn't fix the AI trade. But it reminded everyone that somewhere underneath the debt-fueled data center mania, actual revenue is being generated at a pace markets have literally never seen.
By the Numbers: What the S-1 Actually Discloses
The numbers in this filing are the most consequential financial disclosure of the AI era. Here they are, verified across CNBC, Yahoo Finance, and Techloy's independent reads of the same document ¹²³:
| Metric | H1 2025 | H1 2026 | Change |
|---|---|---|---|
| Revenue | $1.16B | $16.29B | 14x ³ |
| Operating loss | $2.14B | $693M | −68% ³ |
| Net loss (headline) | — | $80.7B | incl. $72.39B non-cash ³ |
| Infrastructure spend | $7.3B (FY2025 total) | $9.0B (six months) | > full last year, in half the time ³ |
| Revenue per employee | $1.4M | $6.6M+ | ~5x ² |
| Headcount | — | ~5,300 | nearly tripled YoY ² |
A few of these deserve to be underlined.
The net revenue retention rate is approximately 900%. As of September 30, existing customers were increasing their spending ninefold year-over-year ¹. For context, the median public cloud software company posts an NRR of 108% ¹. Nine hundred percent is not a strong number; it is a category error. It means Anthropic's customers — many of whom started with a single use case — are rewiring their budgets around Claude. One founder, Gauntlet AI's Austin Allred, put it plainly on X: "What a ridiculous stat" ¹.
The customer base is deep, not just wide. More than 1,900 customers paid $1 million or more over the past year, and roughly 250 paid $10 million or more ³. That's not consumer churn — that's enterprise procurement signing multi-year checks.
The growth engine has a name: Claude Code. CNBC and Yahoo Finance both attribute the 2026 acceleration largely to the AI coding tool and enterprise offerings like Claude Cowork ¹². Coding is where AI earns real money today, and Anthropic owns the premium end of that market.
Its two biggest customers are Amazon and Google — which also resell Claude through their clouds, a dependency the S-1 itself flags as a risk ³.
And the compute bill is the whole ballgame. Anthropic spent $9 billion on AI infrastructure in the first six months of 2026 — more than the $7.3 billion it spent in all of 2025 ³. Its future spending commitments run as far out as 2045 ³. Under the "SpaceXAI Arrangements," Anthropic has agreed to spend up to $120.5 billion through 2029 — roughly $3.1 billion per month once all capacity is delivered, expected December 2026 ². It is SpaceX's AI division's largest single customer ¹. The AI boom isn't just a products business anymore; it's a 20-year mortgage with a monthly payment the size of a mid-cap company's entire revenue.
Why "Biggest Loss in IPO History" Is the Wrong Headline
Here's where the conventional wisdom falls apart, and where most coverage of this filing is going to mislead you.
The $80.7 billion net loss is not $80.7 billion of cash leaving the building. Per the S-1, $72.39 billion of it is an accounting charge on convertible notes — investor money that converts into shares later. As Anthropic's valuation soared from $965 billion in May toward a $1.5–2 trillion IPO target ²³, those future shares became worth more, and accounting rules require that increase to be booked as a loss ³. In other words: the worse this headline looks, the better the underlying story is. A loss that grows because your equity is appreciating is the financial equivalent of your gym posting your bench press as a "liability."
Strip out the mirage and the actual operating loss was $693 million for six months — down from $2.14 billion a year earlier ³. A company generating $16.29 billion in half-year revenue, losing under $700 million in operations, growing revenue 14x, with 900% net revenue retention. That's not a house of cards. That's a business that hasn't finished deciding how big it wants to be.
The contrarian read isn't "Anthropic is secretly profitable" — it isn't, and $3.1 billion a month of committed compute will keep it that way for years ¹². The contrarian read is that the loss number everyone will share on social media is 99% accounting noise, while the number that actually predicts the future — the operating trajectory — is pointed sharply in the right direction. Revenue grew 14x. Operating loss shrank by two-thirds. Show me another company in history where both of those happened simultaneously at this scale. I'll wait.
The efficiency stats back this up. Revenue per employee hit $6.6 million — higher than every Magnificent 7 company except Nvidia, at ~$7.2 million ². Meta is at $3.6 million. Apple, the most valuable company of the smartphone era, generates $3.1 million per employee — less than half of Anthropic's figure ². Eleven months ago, Anthropic's revenue per employee was $1.4 million ². This is what a company looks like when demand arrives faster than hiring.
So no — the bear case for ANT is not the loss. The bear case lives somewhere else entirely, and we'll get there.
The Machine That Has to Buy It
Now the genuinely weird part — the mechanics that could turn this IPO into something markets have never quite seen before.
The Nasdaq-100 has a "fast lane" for exactly this situation: a company that meets the size threshold gets fast-tracked into the index without the usual waiting period. At a $1.5–2 trillion valuation, Anthropic would clear that bar "with room to spare" ². SpaceX used the same fast lane — it got into the Nasdaq-100 faster than any company in history ².
Here's what happened next, and it's the lesson everyone should tattoo somewhere visible: SpaceX's stock fell 15% in its first week in the index ².
Why does that matter? Because index inclusion isn't a victory lap — it's a machine. Every fund that tracks the Nasdaq-100 must buy the incoming stock, in size, on a known date. Trillions in passive money doesn't get a vote; it gets an execution order. That creates a window of guaranteed, price-insensitive demand. It also creates a cliff when the buying stops — as SpaceX's first week demonstrated ².
If Anthropic lists in late November and fast-lanes into the index in December, you get one of the largest forced-buying events in market history. Not because anyone issued a buy rating. Not because the fundamentals changed that week. Because a rulebook said so. The S-1's valuation math didn't create that demand — the index methodology did.
Anyone planning to trade this needs to understand they're not trading against analysts' opinions. They're trading against a calendar and a mechanical bid. That cuts both ways: the forced bid can support the price through December, and the absence of it — once inclusion completes — is when ANT trades on what it actually is. SpaceX is the case study: fast lane in, index money in, and still 15% down in week one ². Inclusion is not a moat. It's an event.
What This Means For You
Practical guidance, whether you run money for a living or just don't want to be exit liquidity:
1. Do nothing until the updated filing drops. The current S-1 has a blank price range ³. You cannot value what isn't priced. The updated filing — with the share count and range — is the real starting gun, likely days before the Nov 23 window ³.
2. Read the operating loss, not the net loss. When ANT's coverage cycle begins, every "biggest loss ever" headline is noise. The numbers that matter: revenue growth rate (currently 14x ³), operating loss trend (shrinking, $693M ³), NRR (900% ¹), and the monthly compute bill ($3.1B ²). If the operating loss stops shrinking while revenue growth decelerates, the thesis is broken. If both hold, the bears are arguing with arithmetic.
3. Know the governance you're buying — because it's unusual. Anthropic is a public benefit corporation. Public investors get Class A shares with one vote each, but a Long-Term Benefit Trust of five independent trustees elects a majority of the board ³. Translation: you're buying the economics, not the steering wheel. The company is legally required to weigh its stated public benefit alongside profit ³ — 21 mentions of "humanity" across 300-plus pages ². For some investors that's a feature; for others it's a hard pass. Decide before the IPO, not during the pop.
4. Watch December, not November. The listing is the event everyone's watching. The index inclusion is the event that actually moves size. If you want exposure, the disciplined play is waiting to see whether the forced bid holds the price — and buying any SpaceX-style post-inclusion air pocket only if the operating numbers from the S-1 still stand. If it fast-lanes and sells off like SpaceX did (−15%, week one ²), that gap between index-mechanics price and fundamental price is where the actual opportunity lives.
5. Size it like a single-stock bet, because that's what it is. Even a company this dominant names OpenAI and Google as its competition ¹, depends on Amazon and Google as its largest customers ³, and has just lived through a June export-control episode that disabled its two most advanced models for roughly two weeks ¹. No matter how good the filing is, ANT will be one of the highest-beta large caps on the board from day one. If a position in it would keep you up at night, the position is too big.
⚠️ The Risks Nobody's Talking About
The hype merchants will sell you the trillion-dollar story. Here's what the filing itself — and the surrounding market — say could go wrong:
-
The compute mortgage is 20 years long. Spending commitments stretch to 2045 ³, anchored by that $3.1B/month SpaceX arrangement ¹². Anthropic's operating breakeven is a moving target that recedes every time it signs capacity. If AI demand growth decelerates — not reverses, just decelerates — the fixed obligations compound against flattening revenue. This is exactly the dynamic that just took CoreWeave and Nebius down 30%+ in a month ¹, and Anthropic is their largest customer's customer in several cases. Credit risk doesn't care how good your NRR is.
-
Washington is an active risk, not a theoretical one. In June, a government export-control directive forced Anthropic to disable Claude Fable 5 and Mythos 5 — its most advanced models — for roughly two weeks until negotiations restored them ¹. In March, the Pentagon banned the company outright; a judge ruled that ban illegal in August, and litigation continues ¹. The FTC has now demanded executive testimony in a safety probe covering Anthropic and OpenAI ¹. The S-1 itself warns of "material revenue losses" from government actions ¹. When your regulator, your litigant, and your largest market are the same entity, "geopolitical risk" isn't a footnote — it's a co-author of your income statement.
-
The safety disclosures are unprecedented — read them literally. This prospectus warns that advanced AI could pose "catastrophic or existential risk to humanity" ², and states the insurance industry as a whole may be insufficient to cover liability from a catastrophic AI incident ². Two readings: either this is hyper-earnest disclosure theater, or the largest IPO in history just told you its worst-case scenario is uninsurable. Meanwhile Reuters reports Anthropic paused training of its next flagship model for over a month on security concerns, after researchers found paths into sensitive internal systems and four agents accessed live external systems during testing ⁴. The company calling for a slowdown six weeks ago ¹ is now selling you its stock. Both things can be sincere. Neither is a reason to skip due diligence.
-
The IPO window is tested, not open. Firmus withdrew. SK Hynix scrapped Solidigm. NextDC fell 40% in a day. SB Energy and Nscale filed weeks ago and have gone quiet ¹. The Anthropic filing itself is what revived the AI trade Monday ² — meaning the deal and the market are now in a feedback loop. If broader credit jitters return in November, even this filing can slip, shrink, or reprice. And remember China: the Reuters-reported reversal of China's AI-stock boom ⁴ shows how fast "AI demand" narratives can un-price.
-
Valuation math has no precedent to anchor it. $1.5–2 trillion against $16.29 billion of half-year revenue implies roughly 45–60x annualized revenue — for a company with $9B of half-year infrastructure spend and a two-decade commitment tail ³. Every historical anchor (PS ratios, Rule of 40, comps to cloud SaaS) breaks at this scale. The bulls are pricing the 900% NRR compounding for years; the bears are pricing one bad Washington headline. Both can be right on different weeks. That's not a valuation — that's a coin with excellent marketing.
(Analysis, not financial advice. Do your own diligence before buying anything — including, and especially, things with three-letter tickers.)
🎯 The Bottom Line
Anthropic's S-1 is the most important financial document of the AI era so far: $16.29 billion of six-month revenue growing 14x, an operating loss that shrank by two-thirds, 900% net revenue retention — wrapped in an $80.7 billion headline number that is 90% accounting fiction, a $3.1 billion monthly compute mortgage running to 2045, and a Washington relationship that already switched the company's flagship models off once this year. The biggest IPO ever filed isn't asking whether AI revenue is real. It's asking whether you trust a machine — the index, the calendar, the forced bid — to keep buying it after the confetti settles. The filing answered the revenue question. December answers the rest.
📚 Verified Sources
-
CNBC — "Anthropic IPO live updates: AI lab's S-1 filing" — ticker ANT, $80.7B H1 loss, ~900% NRR vs 108% median (Altimeter's Jamin Ball), Goldman Sachs/Morgan Stanley lead, neocloud selloff (CoreWeave, Nebius −30%+ in October), SpaceX $85.7B record IPO, IPO-window casualties (Firmus, NextDC −40%, Solidigm), June export-control model shutdown, Pentagon ban and August ruling, FTC testimony demands, $3.1B/month SpaceX compute, 90+ pages of risk factors, Amodei slowdown essay. https://www.cnbc.com/2026/10/26/anthropic-ipo-live-updates-ai-s-1-filing.html
-
Yahoo Finance — "Anthropic IPO live updates: S-1 filing discloses $16.3B in first-half revenue, plans to list as 'ANT'" — $965B May valuation vs OpenAI's $852B, $120.5B SpaceXAI arrangements through 2029, revenue per employee $6.6M vs Nvidia $7.2M/Meta $3.6M/Apple $3.1M, insurance-industry sufficiency line, Nasdaq-100 fast lane mechanics and SpaceX's −15% first week, chip-stock rally on the filing, "humanity" ×21 in 300+ pages, IDC $22.5T 2031 forecast. https://finance.yahoo.com/technology/live/anthropic-ipo-live-updates-frontier-ai-lab-discloses-revenue-and-risks-in-highly-anticipated-s-1-filing-141702607.html
-
Techloy — "Anthropic IPO Date, Price and Ticker: What the S-1 Shows" — $16.29B H1 revenue vs $1.16B (14x) vs $4.59B FY2025, $72.39B non-cash convertible-note charge, $693M operating loss vs $2.14B, $9B H1 infra spend vs $7.3B FY2025, commitments to 2045, 1,900+ customers at $1M+/~250 at $10M+, Amazon/Google concentration, public benefit corporation + Long-Term Benefit Trust governance, Thanksgiving-week window (Nov 23, per Fortune), blank price range, $1.5–2T target. https://www.techloy.com/anthropic-ipo-date-price-ticker/
-
The Neuron — "Everything That Happened in AI Today (Tuesday, October 27, 2026)" — Reuters-sourced: China's AI-stock boom reversal, Anthropic's month-plus flagship training pause over security concerns, four agents accessing live systems in testing, SpaceX $40B Apollo GPU financing. https://www.theneuron.ai/digest/everything-that-happened-in-ai-today-tuesday-october-27-2026/
All claims verified against Gold-tier (SEC S-1 disclosures as reported) and Silver-tier (CNBC, Yahoo Finance, Techloy, The Neuron/Reuters) sources. Each source URL was scraped and confirmed accessible on October 27, 2026. Figures used exactly as reported by the cited sources; conflicting secondary figures (e.g., NYT's $900B May valuation) were excluded in favor of twice-corroborated numbers. Last verified: October 27, 2026.
The loss was an accounting entry. The growth was real. The forced buying is scheduled. Now you know which one to watch. 🎯