SCIENCE & TECH
Editor's note: this article was drafted by an AI model made by Anthropic, the company it covers. Figures come from the S-1 as reported by Yahoo Finance and independent analysts, and have not been audited by this publication.
It is the filing that Silicon Valley, Wall Street and every chief executive with an AI strategy have been waiting for. One week after Anthropic quietly lodged its prospectus with the US Securities and Exchange Commission, the numbers inside are still making people do a double take: $16.3 billion of revenue in six months, an $80.7 billion net loss on the same pages, and a company that says in black and white that its own technology could pose a "catastrophic or existential risk to humanity" if misused or left to run autonomously.
The maker of the Claude chatbot filed its S-1 on Monday 26 October, setting up a Nasdaq listing under the ticker ANT. Goldman Sachs and Morgan Stanley are leading the deal, with a 22-bank syndicate behind them. The price range, share count and size of the raise have been left blank for now, but the chatter on trading floors is already enormous. Reports put the expected valuation anywhere from $1.5 trillion to north of $2 trillion, against the $965 billion the company was valued at in a private round as recently as May.
If the listing goes ahead this fall as planned, it will be one of the largest stock-market debuts in history, and the first time the public can buy a slice of a pure-play frontier AI laboratory. For ordinary investors, pension funds and anyone with a tech-heavy index fund, this is no longer an abstract Silicon Valley story. It is about to touch their money.
The growth number that sounds like a typo
Start with the headline figure. Anthropic reports revenue of roughly $16.3 billion for the first half of 2026. For comparison, Yahoo Finance reports that the whole of 2025 brought in $4.6 billion. In other words, the company booked more than three times last year's total revenue in just two quarters.
The quarterly picture is even starker. Revenue climbed from about $4.7 billion in the first quarter to $11.6 billion in the second. One independent reading of the filing, by The VC Corner, calculates that second-quarter revenue was up 1,368 percent on the same quarter a year earlier, and that first-half revenue compares with only around $1.2 billion in the first half of 2025. Growth like that is almost unheard of for a company of this size. Even the great internet and cloud success stories took years to climb a similar ramp.
The customer data tells the same story. The filing shows the number of customers paying at least $100,000 a year rising from more than 950 to more than 11,900 in twelve months. The VC Corner also highlights a net revenue retention figure of around 900 percent, meaning that, on average, existing customers are spending far more with the company than they did a year ago. Companies are not just trying the technology. They are building it into the way they work, and then spending more and more on it.
The company is also doing all of this with a surprisingly small team. Anthropic had about 5,300 employees at the end of September, and revenue per employee now tops $6.6 million, up from $1.4 million a year earlier, according to Yahoo Finance's reading of the document.
So why is there an $80 billion loss?
Here is where the story gets messy, and where headline writers are having a field day. The same filing reports a net loss of $80.7 billion for the first half of 2026, up from $42 billion for all of 2025.
But this is not quite what it looks like. According to the VC Corner's analysis, around $72.4 billion of that loss is non-cash: accounting adjustments tied to changes in the fair value of convertible notes, derivatives and SAFEs, the investment instruments startups use to raise money before they go public. In plain English, a big part of the red ink is a paper charge that reflects how much more the company is now worth, not cash flowing out of the door. Strip it away and the underlying picture looks far healthier.
Indeed, the operating numbers show a turning point. Anthropic booked an operating profit of $974 million in the second quarter, after an operating loss of $1.7 billion in the first. It is the first profitable operating quarter in the six quarters shown in the filing. Over the same period, compute costs rose from $3.4 billion to $5.6 billion, so revenue is growing far faster than the biggest cost on the books.
Treat the caveat with care, though. These breakdowns come from secondary readings of a complex document, and the final prospectus will be picked apart line by line by auditors, regulators and short sellers in the weeks ahead.
The computing bill that could make your eyes water
Behind every chatbot answer sits a data centre humming with expensive chips, and Anthropic's filing shows just how much it has promised to spend on them. Under what the document calls its "SpaceXAI Arrangements", the company has agreed to spend up to $120.5 billion through 2029. That works out to roughly $3.1 billion a month once the capacity is fully delivered, which Anthropic expects to happen in December 2026.
That is only one contract. The VC Corner adds up named compute agreements with Google, Amazon, Microsoft, SpaceXAI, Broadcom, Nscale and Lambda and arrives at roughly $637 billion in total commitments, a figure the author calculated from pages of the filing and which has not been confirmed by the company. Anthropic is also shifting toward a hybrid model it calls "Anthropic Direct", under which it leases and operates its own data centres alongside the cloud services it buys from Amazon and Google.
Against all of that, the company reports cash and short-term investments of $131.7 billion. It is a gigantic war chest, but also a sign of the scale of the bet: the AI industry is wagering that demand will keep growing fast enough to justify building computing capacity measured in the hundreds of billions.
Who actually controls the company?
Investors buying into ANT will not be getting a conventional say in the running of the business. According to the VC Corner's reading of the filing, seven co-founders will hold 50.1 percent of the voting power through a single Class F share held by a founder vehicle. On top of that, Anthropic's Long-Term Benefit Trust, an independent body created to keep the company's safety mission in view, elects a majority of the board.
Anthropic is also structured as a public benefit corporation, meaning it is legally allowed, and expected, to weigh its stated public mission alongside profit. Supporters say that is a safeguard. Critics say it may leave ordinary shareholders with little power if their interests clash with the founders' or the trust's.
Another unusual wrinkle is the lockup. Instead of the standard 180-day wait before insiders can sell, Yahoo Finance reports that some can sell from the first day of trading, with the remainder unlocking in more than 20 staggered windows spread over about a year. Supporters say it avoids a single cliff that could crash the share price. Sceptics will ask what it means for confidence if insiders can cash out at once.
The risk section nobody can ignore
Most IPO risk factors are dry legal boilerplate. This one stands out. Anthropic's prospectus warns that AI models might pose a "catastrophic or existential risk to humanity" if misused or allowed to operate autonomously. It flags "increasingly restrictive AI regulation" as a threat to the business. And, as AI researcher Dave Banerjee pointed out on X, it concedes that the insurance industry as a whole may not be big enough to cover the company's liability in a catastrophic AI incident.
Whether that is admirable honesty or a lawyer's insurance policy depends on who you ask. What is clear is that investors are being asked to buy shares in a company that says out loud that its product line could, in the worst case, cause damage nobody can insure.
The arms race around the listing
The timing is no accident. The AI industry has barely paused for breath in recent weeks. OpenAI launched GPT-6.1 Terra on 26 October, claiming performance close to its top-tier model at a tenth of the cost. Google unveiled Gemini 4 Carbon on 28 October, its most powerful model yet, though for now it is limited to a select group of cyber-defence experts. French lab Mistral released the open weights of a roughly one-trillion-parameter model the next day. All of these performance claims are the companies' own.
Anthropic's biggest rival, OpenAI, is reportedly targeting a $1.25 trillion valuation in a pre-IPO funding round and is expected to hold off on its own stock-market debut until early 2027. That would give Anthropic the chance to be the first major frontier lab to open its books to public scrutiny, and the first to learn how the market prices the business.
The market reaction was immediate. Chip stocks including Nvidia, Broadcom, Micron and Intel rose after the filing, helping to lift the Nasdaq, a sign that investors see Anthropic's spending as good news for the companies that sell the hardware.
Expert perspective: dazzled, but watching the fine print
Reaction among analysts has been a mixture of awe and caution. Those impressed by the filing point to the growth, the shift to operating profit and the retention figures as evidence that enterprise demand for AI is real, not hype. Those wary of it point to the sheer size of the spending commitments, the dependence on a handful of giant suppliers, and the fact that most of the performance claims in the AI industry are self-reported.
Independent commentators who have gone through the document, including the analysts at The VC Corner and Yahoo Finance's live coverage team, have been careful to separate the audited facts from the estimates. The largest numbers, such as the $637 billion in total compute commitments and the non-cash share of the loss, are readings of the filing rather than statements from the company, and the blank run-rate revenue line on one page of the filing means some of the most closely watched data point is still missing.
The sensible view for now is that the filing is a snapshot of a company growing at breakneck speed and spending at an even more breakneck pace. Whether the two curves keep running together is the multi-trillion-dollar question.
What this means for you
If you invest, even indirectly, through a tracker fund or a workplace pension, the listing could soon put a very large new stock in the index you own. If you use Claude for work, the filing suggests that the tool you rely on is backed by a company with more than $130 billion in cash, which is reassuring, but also one that is spending money as fast as it can bring it in.
If you work in an industry being reshaped by AI, the revenue figures will be a reminder that big employers are paying real money for these systems, and that they are doing so now. And if you simply watch the technology with a mix of fascination and unease, the risk section is a rare case of a company writing down its own worst fears in an official document that investors, regulators and the public can all read.
There is also a warning. A company valued at up to $2 trillion needs to keep growing for years to justify the price. Any stumble, whether a regulatory crackdown, a rival's breakthrough or a hiccup in the supply of chips, could send the stock swinging wildly.
The bottom line
Strip away the dramatic headline numbers and the story is simple. Anthropic's S-1 shows a business that has gone from promising start-up to one of the fastest-growing companies in corporate history in just a few years, while betting hundreds of billions of dollars on a future in which the demand for artificial intelligence keeps growing. The loss figure grabs attention, but the operating profit, the customer growth and the cash pile may matter more.
The final word will come when the price range is published, the roadshow begins and the first trade flashes up on the Nasdaq screen under the letters ANT. Until then, the filing has done something rare: it has turned the AI boom from a story about demos and promises into one about spreadsheets, and the spreadsheets are astonishing.
This is a developing story. Nothing here is investment advice, and readers should consult the filing itself on the SEC's EDGAR database before making any financial decision.
