Photo: Pexels
On Monday, October 26, 2026, the most anticipated stock market debut in artificial intelligence stopped being a rumour and became a document. Anthropic, the company behind the Claude family of AI models, publicly filed its S-1 prospectus with the US Securities and Exchange Commission, applying to list its Class A shares on the Nasdaq Global Select Market under the ticker symbol "ANT". Goldman Sachs and Morgan Stanley are leading a syndicate of more than twenty banks.
That single filing did something months of leaks could not: it replaced speculation with audited numbers. For the first time, anyone can read exactly how much money the world's most closely watched private AI company makes, what it spends, and who will still be in charge after ordinary investors are allowed to buy in. The headline figures look contradictory at first glance — $16.3 billion of revenue in six months, alongside an $80.7 billion loss — and that contradiction is precisely why this filing matters. It is the clearest public test yet of whether the AI boom is a real business or an expensive story told beautifully.
This article explains what Anthropic actually filed, what the numbers really mean, why the company's founders are keeping control, and what happens next — including why the same day's $95 billion fundraising by Meta tells a second, equally important part of the story.
What Anthropic filed, and what is still blank
An S-1 is the document a company files when it wants to sell shares to the public for the first time. It is part prospectus, part confession: the company must lay out its finances, its risks, its customers, and its governance in a form the SEC can scrutinise and investors can sue over if it misleads them. Anthropic had filed confidentially in June 2026, and reporting through the summer moved the expected public filing several times — first late September or early October, then mid-October at the earliest. The public version finally arrived on October 26, roughly a month behind the earliest targets.
Two things investors most want to know are still missing, and the filing says so plainly: the number of shares to be offered and the price range have not yet been determined. The registration is also not yet effective, which means no shares can be sold and no offers to buy can be accepted until the SEC declares it effective. A later amendment, known as an S-1/A, will fill in those blanks, and reporting has pointed to a roadshow — the tour on which bankers pitch the deal to big investors — in the week of November 9. Until that pricing happens, every valuation figure attached to Anthropic, including Reuters' reporting that the IPO could value the company at more than $2 trillion, is a press report, not a filed fact.
The revenue number that changed the conversation
The filing shows revenue of $16,286 million in the first six months of 2026, against $1,163 million in the same period of 2025. That is roughly a fourteen-fold jump in a year. Full-year revenue was $4,585 million in 2025, $386 million in 2024, and $41 million in 2023, the year Anthropic launched commercially. The second quarter of 2026 alone brought in $11.6 billion.
That trajectory is why Wall Street paid attention. Very few companies in history have gone from $41 million to a $16 billion half-year in three years, and none of them did it selling a product — large language models — that barely existed as a commercial category a decade ago. The growth also reframes the debate about AI spending: Anthropic is not only a consumer of computing power, it is now one of the largest revenue-generating software businesses in the world, built almost entirely on selling access to Claude through application programming interfaces and cloud partnerships with Amazon, Google and Microsoft.
There is a concentration warning buried in the same section, and it deserves attention. Approximately 20 percent of first-half 2026 revenue came from just two unnamed customers, according to The Information's reading of the filing, down from nearly a quarter of 2025 revenue. The filing does not name the two accounts. For a company of this size, that is not unusual — enterprise software often starts concentrated — but it means a single lost contract or a single renegotiated cloud deal can move the numbers visibly. Any investor reading the filing closely will be asking who those two customers are and how long their contracts run.
The $80.7 billion loss that is not what it looks like
The figure that travelled fastest around the world on Monday was the net loss: $80,748 million for the first half of 2026, up from $41,965 million for all of 2025, and following net losses of $1.5 billion in 2023 and $8.3 billion in 2024. Taken at face value, it sounds like a company burning money at a pace no business could survive.
It is not what it looks like, and the filing itself explains why. Roughly $72.4 billion of the half-year loss is a non-cash charge from changes in the fair value of convertible notes, derivative liabilities and SAFEs — instruments that promise investors future shares. When Anthropic's own valuation rises, the accounting value of those promises rises too, and the increase is recorded as a loss. In other words, the company looks worse on paper precisely because the market thinks it is worth more. It is an accounting side-effect of success, not cash leaving the bank.
The cleaner measure is the operating result, which strips out that paper charge. Anthropic's operating loss for the first half of 2026 was $693 million, against $2.14 billion in the first half of 2025 and $8.06 billion for all of 2025. More strikingly, the second quarter of 2026 produced operating income of $974 million — about $1.0 billion, an 8 percent operating margin — the first quarterly operating profit in the company's history. Operating cash flow for the half-year was $3.93 billion positive. A company can lose $80 billion on paper and still generate cash, and Anthropic's filing is now the textbook example.
That does not make the business cheap to run. Infrastructure spending climbed to $8.95 billion in the half-year, from $2.33 billion a year earlier, as Anthropic bought the computing capacity its models need. TradingView's summary of the filing also put cash and short-term investments at $131.65 billion. The honest summary is therefore a mixed one: a business growing at extraordinary speed, turning operationally profitable in one quarter, but with a cost base that roughly quadrupled in a year and will keep rising as long as the model race continues.
Who keeps control after the IPO
Public buyers of ANT shares will get the economics of the company, but not control of it. That is deliberate, and the filing spells out the mechanism in unusual detail.
Anthropic is incorporated as a public benefit corporation under Delaware law. A single Class F share, held by a vehicle controlled by the seven co-founders, will carry 50.1 percent of the general voting power. Separately, the company's Long-Term Benefit Trust will generally have the authority to elect and remove the Class T directors, who will make up a majority of the board. Each Class A share bought by the public carries one vote, but that vote has limited reach, since the trust and the founders decide most outcomes. Anthropic says it is a "controlled company" under Nasdaq's corporate governance rules.
This structure continues a philosophy Anthropic set out earlier in October, when it created a Public Benefit Fund holding 5 percent of the company, with its stake rising as the company's value rises. The message to investors is consistent: Anthropic is willing to take public money, but it is not willing to let public shareholders redirect its mission — which, in the company's own framing, includes managing the risks of the very technology it sells.
The risk section written about its own product
Every prospectus contains a risk-factors section listing what could go wrong. Anthropic's is different from most because it is written by a frontier AI developer about its own core product. Reuters reported before the filing that Anthropic planned to warn prospective investors that advanced AI could pose catastrophic or existential risks to humanity, and that warning language now sits in the same document as the revenue figures, in front of the institutional buyers who will decide the IPO price.
Risk language in a prospectus is partly legal protection, and no investor should read it as a prediction. But its presence matters for a different reason: it forces the public conversation about AI safety into a disclosure document with legal weight, rather than leaving it to blog posts and conference panels. If Anthropic's models cause harm, or if regulation constrains them, the company cannot later say investors were never told. For regulators, journalists and researchers, the risk section will become a reference text in its own right, quoted back at the company for years.
The same day Meta asked for $95 billion
Anthropic's filing did not arrive alone. On the same Monday, Meta Platforms said it plans to raise up to $95 billion in its largest-ever equity fundraising to pay for AI data centres and computing power. The plan has two parts: about $35 billion in public offerings now — roughly $22.5 billion of Class A common stock sold through forward sale agreements and about $12.5 billion of mandatory convertible preferred stock, sold as 250 million depositary shares Meta plans to list on Nasdaq as "METAM" — and a separate at-the-market programme of up to $60 billion once its new non-voting Class C shares begin trading, expected in the first quarter of 2027.
Meta's numbers explain why even a company with $82.45 billion in cash, cash equivalents and marketable securities at September 30, against $80.92 billion of long-term debt, still needs to sell shares. Its capital spending guide for 2026 is $135 billion to $145 billion, and Business Times reported the company has committed to spend about $825 billion in the future, primarily on AI infrastructure and cloud computing. Nvidia, AMD and Broadcom have expressed interest in buying up to $5 billion of Meta's Class A shares in the offering — chip suppliers financing one of their biggest customers, a circular flow of capital that analysts at William Blair flagged, saying investors will keep questioning the financing and the potential for future raises.
The market's split verdict on Monday captured the tension neatly. Wall Street closed higher, with the S&P 500 up 0.70 percent to 7,676.77, the Nasdaq up 0.83 percent to 26,169.38 and the Dow up 0.28 percent to 52,361.58, as Anthropic's filing revived optimism about AI stocks; the Philadelphia Semiconductor Index rose 2.8 percent, though it remained down 12.2 percent for the month, and AI-linked names Lumentum and Coherent gained about 8 percent and 6 percent. Meta itself fell 1.9 percent. Investors, in other words, cheered the company going public and punished the company paying for the privilege of competing with it.
What this filing does — and does not — settle about an AI bubble
Searches for "ai bubble" in Pakistan rose more than 1,000 percent in the past day, according to Google Trends, and the question behind them is fair: if companies must raise tens of billions and file $80 billion paper losses, is this a bubble? Anthropic's filing gives both sides better evidence than they had last week, and that is its real value.
The bubble case now has to reckon with $16.3 billion of half-year revenue, a profitable quarter and positive operating cash flow. Those are not the numbers of a company selling only a story. The sceptical case, meanwhile, has sharper material too: infrastructure costs that quadrupled in a year, two customers providing a fifth of revenue, a valuation conversation running into the trillions before a single share has been priced, and a funding model in which chipmakers buy stock in their customers. Neither side can now claim the other has no evidence.
The most useful way to read the filing is as a hinge, not a verdict. Before October 26, the AI industry's economics were argued from leaks, private funding rounds and executive interviews. From now on, at least one leading company's economics are argued from a public document, updated every quarter, under securities law. That transparency will discipline both the hype and the panic, because every future claim about Anthropic's business can be checked against a number it filed itself.
Three things will decide whether this IPO is remembered as a triumph or a warning. First, the pricing: the roadshow reported for the week of November 9 will show what institutional investors will actually pay, not what headlines suggest. Second, the next two quarters: one profitable quarter is a data point, not a trend, and the filing's cost curve gives sceptics a clear test. Third, the earnings of the giants reporting this week — Alphabet, Microsoft and Amazon — alongside the Federal Reserve's interest-rate decision on October 28, will show whether the customers and cloud partners funding this boom are themselves converting AI spending into profit. Anthropic's filing opened the books. The market will now do the auditing.
Frequently asked questions
What is Anthropic's IPO ticker symbol and exchange?
Anthropic has applied to list its Class A common stock on the Nasdaq Global Select Market under the ticker symbol "ANT". The shares are not trading yet, because the registration statement is not yet effective and the share count and price range have not been set.
How can Anthropic report an $80.7 billion loss and still be valuable?
About $72.4 billion of the first-half 2026 net loss is a non-cash accounting charge from revaluing convertible notes and similar instruments as Anthropic's own valuation rose. The operating loss, which better reflects the running business, was $693 million for the half-year, and the second quarter was operationally profitable.
Did Anthropic actually make a profit?
Yes, on an operating basis, in one quarter: the second quarter of 2026 produced $974 million of operating income on $11.6 billion of revenue, an 8 percent operating margin. For the full half-year, however, Anthropic still recorded an operating loss of $693 million because the first quarter was loss-making.
Who will control Anthropic after the IPO?
The seven co-founders, through a single Class F share, will hold 50.1 percent of the general voting power, and the Long-Term Benefit Trust will elect a majority of the board. Public Class A shareholders get one vote per share, but the company will be a Nasdaq "controlled company", so public investors will not control its direction.
When can ordinary investors buy ANT shares?
Not until the SEC declares the registration effective and the offering is priced, which happens after the roadshow. The filing gives no date for either step, and no shares can legally be sold before effectiveness, so any advertisement offering ANT shares today should be treated with suspicion.
Does the Anthropic IPO prove AI is not a bubble?
It proves one leading AI company has very large, fast-growing, partly profitable revenues — which weakens the claim that the whole boom is story without sales. It does not settle the bubble question, because infrastructure costs, customer concentration and trillion-dollar valuation talk remain real risks. The filing gives both sides evidence, and the next few quarters of filed results will matter more than the filing itself.
In short: Anthropic's S-1 turns the AI debate from rumour into arithmetic. Revenue of $16.3 billion in six months, a first profitable quarter, a paper loss that grows when the company is worth more, and founders who keep control — that is the deal investors are being asked to price, and the rest of the industry will be judged against it.
For more analysis like this, visit abdulrahimmughal.com.