I have read a lot of IPO filings in my time as a developer, usually only when a tool I depend on is about to change hands. Most of them are boring in a reassuring way: steady revenue, modest losses, a plan. Anthropic's filing, published on October 26, 2026, is not that kind of document. It is the first time a frontier AI lab has laid its finances bare in a public S-1, and the numbers are so large on both sides of the ledger that the headlines almost miss the point.
Anthropic, the company behind Claude, filed a registration statement on Form S-1 with the US Securities and Exchange Commission for a proposed initial public offering of Class A common stock. It plans to list on the Nasdaq Global Select Market under the ticker symbol ANT, with Goldman Sachs and Morgan Stanley leading the offering alongside about twenty other banks. The company first filed confidentially in June 2026. What it does not contain is just as important as what it does: no share count, no price range, and no valuation. The registration statement has not yet become effective, which means no shares can be sold yet.
Press reports have filled that gap with their own numbers — a roadshow possibly starting the week of November 9, trading possibly in the week of November 16, before Thanksgiving on November 26, a raise of up to $100 billion, and a valuation somewhere between $1.5 trillion and $2 trillion. I am treating all of those as reports, not facts, because none of them come from the filing itself.
The Revenue Growth Is Real, and It Is Fast
Let us start with the number that made everyone sit up. Anthropic reported revenue of $16.3 billion for the first six months of 2026, compared with just over $1 billion in the same period a year earlier. For the whole of 2025, revenue was $4.6 billion. In other words, the company made more than three times its entire 2025 revenue in half a year.
The quarterly split makes the acceleration clearer. Revenue rose from $376 million in the first quarter of 2025 to $11.6 billion in the second quarter of 2026. That second quarter also produced Anthropic's first positive GAAP operating profit, $974 million, and for the first half as a whole the operating loss narrowed to $693 million, down from $2.1 billion in the first half of 2025. Adjusted operating income for the half was $1.6 billion, against an adjusted operating loss of $1.8 billion a year earlier.
As someone who pays for AI tools and APIs, I read that operating trend as the most encouraging line in the whole filing. A company whose core operations are approaching break-even is in a very different position from one that loses money on every query it serves. It also helps explain why so much of the revenue is usage-based: around 80 percent of revenue comes from customers paying for how much they use the models, with fixed subscriptions making up the rest. When developers and businesses build real workflows on Claude — coding assistants, customer support, document processing — usage compounds in a way that a flat subscription never does.
There are two caveats I do not want to skip. First, the S-1 does not include third-quarter results, even though the company reportedly already has those figures. Second, roughly a fifth of first-half revenue came from just two customers. That is concentration risk in plain sight.
The $80.7 Billion Loss That Is Not What It Looks Like
Now the number that dominated the headlines: a net loss of $80.7 billion for the first half of 2026, up from about $13 billion in the same period a year earlier. If you stopped there, you would think Anthropic was burning cash at an unimaginable rate. It is not, and understanding why matters if you want to read any AI company's finances honestly.
Almost the entire loss is accounting, not cash. Fair value charges on convertible notes, SAFEs, and note purchase agreements totalled $81.8 billion in the half. Here is the plain-English version: when a fast-growing private company raises money through instruments that convert into shares later, accountants have to revalue those instruments as the company's value rises. The more valuable Anthropic becomes, the bigger the paper charge. It is a real accounting loss, and it will matter to shareholders, but it is not $80.7 billion walking out of the bank account. The operating figures above — a $693 million operating loss on $16.3 billion of revenue — are a much better guide to how the day-to-day business is actually performing.
That distinction is worth keeping in mind for the next few weeks of coverage, because the $80.7 billion figure will be quoted without context in a hundred articles. When a headline tells you a company "lost" eighty billion dollars, ask whether it lost cash or recorded a revaluation. In this case, it is overwhelmingly the second.
The Compute Bill: Hundreds of Billions, Committed in Advance
If the loss is less scary than it looks, the compute commitments are more serious than they sound — and this is the part of the filing I keep coming back to as a developer.
Anthropic spent $8.9 billion on computing power and infrastructure in the first half of 2026, an increase of about 283 percent from a year earlier, and more than it spent in all of 2025. Yet compute costs as a share of revenue have been falling fast: they exceeded ten times revenue in 2023, fell to 653 percent in 2024 and 160 percent in 2025, and stood at 55 percent in the first half of 2026. Revenue is growing faster than the compute bill, which is exactly what has to happen for this business model to work.
The future commitments, though, are on another scale entirely. Since the end of 2025, Anthropic has entered into additional computing agreements with minimum payment obligations of approximately $417 billion. Those contracts are generally expected to begin in 2026 or 2027 and can run for as long as twenty years. You may also see much larger totals quoted — $780 billion, $810 billion, even $855 billion — and that is not a contradiction. Different reports count different things: minimum obligations versus total commitments, cloud agreements versus equipment leases, cancellable versus non-cancellable deals. The filing names at least $110 billion with Amazon, $114.5 billion with Google, and $44.9 billion with Microsoft, plus about $161.2 billion in rent on Broadcom equipment leases and up to $120.5 billion with SpaceXAI, the last of which can be cancelled on ninety days' notice after initial periods ending in early 2027. The Amazon and Google commitments, by contrast, apply regardless of actual usage: if Anthropic spends less, it pays the difference.
Control Stays With the Founders
One more part of the filing deserves attention, because it shapes how much say public shareholders will actually have. Anthropic is organised as a public benefit corporation, and the S-1 describes a Founder LLC, initially made up solely of the seven co-founders — Dario Amodei, Daniela Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish, and Christopher Olah — which will receive a single share of Class F common stock immediately before the offering completes. That share carries votes equal to 50.1 percent of the total voting power of Anthropic's outstanding capital stock.
What This Means for Developers Like Me
So what should a working developer actually take from a 300-page prospectus? I would boil it down to four things.
First, AI coding and API pricing are now tied to public-market economics in a way they were not a year ago. When Anthropic was private, its pricing decisions were its own business. As a public company with quarterly reporting, every price change, every new model tier, and every usage trend will be scrutinised. That scrutiny can be good for users — it rewards efficiency — but it also means the era of heavily subsidised API access may not last forever.
Second, competition is about to get more visible. OpenAI has reportedly delayed its own IPO plans, with its chief executive said to have called 2026 an ill-advised moment for a public offering. If Anthropic lists first, it becomes the public benchmark for the whole frontier-lab business model: its revenue, its margins, and its compute bill will be the numbers every rival gets judged against. For those of us choosing between providers, that transparency is genuinely useful.
Third, do not ignore the concentration and commitment risks, even if you are not buying shares. If a fifth of Anthropic's revenue comes from two customers, and hundreds of billions are committed to a handful of suppliers, the Claude ecosystem depends on a small number of very large relationships. That is not a reason to avoid building on Claude, but it is a reason to keep your architecture portable: abstract your model calls and keep prompts provider-neutral where you can, so a pricing change or a provider pivot does not break what you built.
Fourth, watch the roadshow, not just the listing. The amended filing with a price range will be the first real signal of what the market thinks this company is worth. The questions institutional investors ask about compute costs, customer concentration, and the path to sustained operating profit are the same questions developers should ask about any platform they build a business on.
Conclusion
Anthropic's S-1 is a strange document: $16.3 billion of revenue in six months, an $80.7 billion net loss that is mostly paper, $8.9 billion of compute spending in a half-year, and commitments running into the hundreds of billions. Read carefully, it looks like a company whose operating business is approaching profitability while betting its future on infrastructure at a scale few businesses have ever attempted.
Frequently Asked Questions
When did Anthropic file for its IPO? Anthropic filed a Form S-1 registration statement with the US SEC on October 26, 2026, after filing confidentially in June 2026. It plans to list Class A common stock on the Nasdaq Global Select Market under the ticker ANT.
Has Anthropic set an IPO price or date? No. The filing leaves the share count and price range blank, and the registration statement has not yet become effective, so no shares can be sold yet. Press reports suggest a roadshow in the week of November 9 and possible trading in the week of November 16, 2026, but Anthropic has not confirmed a date.
How much revenue did Anthropic report? Revenue was $16.3 billion in the first half of 2026, compared with just over $1 billion in the same period a year earlier and $4.6 billion for all of 2025. Second-quarter revenue alone was $11.6 billion, with a first positive GAAP operating profit of $974 million.
Why is the $80.7 billion net loss misleading? Most of it is non-cash accounting. Fair value charges on convertible notes, SAFEs, and note purchase agreements totalled $81.8 billion in the half. The operating loss, a better guide to day-to-day performance, narrowed to $693 million on $16.3 billion of revenue.
How large are Anthropic's compute commitments? Since the end of 2025 it has entered into additional agreements with minimum payment obligations of about $417 billion, with contracts running up to twenty years. Broader tallies that include leases and other deals reach roughly $780 billion to $855 billion, depending on what is counted. Named commitments include at least $110 billion with Amazon, $114.5 billion with Google, and $44.9 billion with Microsoft.
Will the founders keep control after the IPO? Yes, under the proposed structure. A Founder LLC made up initially of the seven co-founders will hold a Class F share carrying 50.1 percent of total voting power, so the founders will be able to determine the outcome of shareholder votes for the foreseeable future, if the offering completes as described.
