The $80 Billion Loss Is the Distraction. The Contract Is the Story.
Anthropic filed to go public, and the filing reads less like a coming-out party than a procurement ledger: staggering revenue, a paper loss, and half a trillion dollars of compute already promised.
The interesting thing about Anthropic’s S-1 is not the $80.7 billion loss. It is the fact that the loss is mostly not real, while the bill behind it very much is.
Anthropic publicly filed its S-1 with the SEC on Monday, October 26, 2026, to list Class A shares on the Nasdaq Global Select Market under the ticker ANT. The price range and the share count are blank. Goldman Sachs and Morgan Stanley lead a 24-bank underwriting group, according to the company’s announcement and filing coverage. What the filing does contain, in unusual detail, is the economics of running a frontier lab in 2026 — and those economics are the story.
A revenue curve with no precedent, and a loss that isn’t cash
Start with the top line, because it is genuinely without precedent for a company this young. Anthropic’s revenue was $41 million in 2023, $386 million in 2024, and $4.59 billion in 2025, according to the filing. The first half of 2026 alone brought in $16.29 billion, against $1.16 billion in the same period a year earlier — growth of roughly 1,300 percent. The June quarter alone was $11.56 billion, up 1,368 percent year over year.
Then the headline number: a net loss of $80.7 billion for the same six months. That sounds like a company on fire. It isn’t, quite. About $72.4 billion of it is a non-cash accounting charge — the paper value of convertible notes, derivative liabilities, and SAFEs rising as Anthropic’s own valuation rose. Accounting rules book that increase as a loss. Strip it out and the operating picture is far more ordinary: a $693 million operating loss for the half, and a $974 million operating profit in the second quarter alone, the company’s first profitable operating quarter, with positive operating cash flow of $3.93 billion.
That sounds like a company that has crossed into viability. It hasn’t, quite, either — and that is where the filing gets interesting.
The real bill: compute, committed in advance
Anthropic spent nearly $9 billion on computing capacity and infrastructure in the first half of 2026, according to filing summaries — already more than the $7.3 billion it spent in all of 2025. And that is the spending that already happened. The spending that is promised dwarfs it: summaries of the filing put future compute and infrastructure commitments above $518 billion over roughly the next decade, with about 80 percent of them non-cancellable.
The named deals alone tell the shape of the industry. Roughly $111 billion with Google, roughly $110 billion with Amazon Web Services, and roughly $31 billion with Microsoft, plus equipment leases tied to Broadcom reported at around $161 billion. One tally of the filing’s named contracts reaches roughly $637 billion. The exact total depends on which agreements you count and how you treat options and extensions — the filing’s own presentation is what the summaries are reading — but the direction is not in doubt: a frontier lab is now, in large part, a vehicle for converting investor capital into long-dated cloud contracts.
That is also why the customer concentration matters. Two customers accounted for about one-fifth of first-half revenue, according to Wall Street Journal reporting on the filing. Against that, the breadth is real too: 11,900 customers generating at least $100,000 in annual revenue as of September 30, 2026, 1,900 at $1 million or more, and 250 at $10 million or more, per the filing. The base is widening fast. It is still narrow at the top.
The Hacker News discussion, which put the filing at 716 points and 649 comments on the day it appeared, treated the $80.7 billion as the punchline. That is the wrong punchline. A paper loss created by your own rising valuation is a curiosity. A decade of non-cancellable compute is a strategy — and a bet that demand keeps growing fast enough to fill capacity that has not been built yet.
You can buy the stock. You cannot buy a say.
The second thing the filing makes plain is that public investors are being offered economics without control, by design. Anthropic is a Delaware public benefit corporation. Just before the offering closes, a single Class F share goes to a company owned by the seven co-founders, and that one share carries 50.1 percent of the vote on ordinary shareholder matters. It has no economic rights. Separately, a Long-Term Benefit Trust holds a Class T share that lets it elect directors making up a majority of the board. Anthropic will be a controlled company under Nasdaq rules.
There is also a Public Benefit Fund holding 5 percent of the company, with more shares added as the price rises, and each co-founder has pledged to give 80 percent of personal Anthropic equity to philanthropic causes. You can read that as mission protection or as insulation from accountability, and the filing does not settle which. What it settles is the mechanics: the people who built the lab keep steering it, whatever the ticker does.
Then there is the price nobody has set. Anthropic’s last private round, in May 2026, raised $65 billion at a $965 billion valuation. Reporting puts IPO expectations between roughly $1.5 trillion and $2 trillion, with a raise of up to $100 billion — but the filing itself sets no price, and those figures are bankers’ and investors’ talk, not offering terms. Anthropic separately told investors that third-quarter revenue reached $16.8 billion with adjusted profitability, according to October reporting — but note the boundary: the filed financials run only through June 30, and the third-quarter figure is investor communication, not part of the S-1.
For builders, the takeaway is practical
For builders, the takeaway is practical: the frontier business model has converged on something simple. Revenue grows faster than almost anything in business history, and it is immediately converted into capacity commitments that last longer than most model generations. The token prices you pay are downstream of those contracts. When a lab signs a ten-year, non-cancellable cloud deal, it is not buying computers. It is buying the right to keep growing at a rate no one can promise.
That sounds like scale. It isn’t, quite — it is leverage. If demand keeps compounding, the contracts look cheap. If it stalls, they are still due. Anthropic’s filing shows a company that has already survived the first version of that bet: revenue compounding, an operating profit in the June quarter, cash flow positive. The IPO asks public markets to fund the second, much larger version.
There is a version of this story in which the S-1 is a triumph: the lab that proved a frontier model business can be real, filing to let the public in. There is another version in which it is a warning label, printed in 335 pages: the largest numbers in the document are not the revenue or the loss, but the promises. The loss will fade from the accounts as the old instruments convert. The contracts will not fade for a decade.
That is the problem with reading an S-1 as a scoreboard. The score that matters is the one Anthropic has already committed to paying, whether the future cooperates or not.
