Anthropic’s S-1 Explained: $17B Revenue, a $567M Loss and an $895B Bill

By Rana — Editor, Elite Era Trends

Anthropic, the company behind the Claude chatbot, filed the paperwork to sell shares to the public on October 26, 2026. A week later, on November 2, it filed an amended version that added something investors had been waiting for: a full quarter of financial results. The numbers inside are startling in both directions. Revenue is growing at a speed few companies have ever matched. The bill for the computing power behind that growth is growing even faster.

This is a reading guide, not investment advice. An S-1 is a long, technical document, and the headlines it produces tend to pick one number and run with it — usually the biggest. Below, the filing’s key figures are set out in plain English, in the order a non-investor actually needs them: what Anthropic sells, how fast it is growing, why it swung back to a loss, what the $895 billion really is, who controls the company after the listing, and what is still blank. Every figure comes from the filing itself, as published on the SEC’s EDGAR database.

Contents

• The short answer
• What is an S-1, and what did Anthropic file?
• The numbers that matter
• Why a $17 billion quarter still lost money
• The $895 billion question
• Who will actually control Anthropic
• The risk factor nobody expected
• What happens next
• Frequently asked questions
• Sources

The short answer

Anthropic’s amended S-1, filed November 2, 2026, reports $16.96 billion of revenue for the three months ended September 30, 2026 — up from $1.38 billion a year earlier — and a $567 million operating loss, after a $974 million operating profit the quarter before. The headline $23.29 billion net loss is mostly a non-cash accounting charge on convertible notes, not money spent running the business. The filing also discloses about $895.4 billion in compute and infrastructure commitments, and it still leaves the share price, the number of shares and the listing date blank. The company has applied to list Class A shares on the Nasdaq under the ticker ANT, but there is nothing to buy yet, and no price has been set.

What is an S-1, and what did Anthropic file?

An S-1 is the registration statement a private company must publish before it can sell shares to the public in the United States. It is filed with the Securities and Exchange Commission (SEC) and published in full on the SEC’s EDGAR database, where anyone can read it. It covers the business, the financial statements, the risks, how the money will be used, and who controls the company.

Anthropic’s timeline so far:

• June 1, 2026: Anthropic confidentially submitted a draft S-1 to the SEC.
• October 26, 2026: the company published its first public S-1. It covered financials only through June 30, 2026: revenue of $16.29 billion for the first half of the year, against $1.16 billion a year earlier, and a net loss of $80.75 billion.
• November 2, 2026: Anthropic filed Amendment No. 1, known as an S-1/A. It added the third quarter’s results, refreshed the risk disclosures, and filled in the total of the company’s spending commitments, which the first filing had left blank.

The amendment is the document this guide reads. Its registration number is 333-299633, and the full text is on EDGAR. Two things are still missing from it: the price range for the shares and the number of shares to be sold. A further amendment is expected to fill those in before any listing, and the registration statement is not yet effective — the offering cannot be priced until the SEC declares it so.

The numbers that matter

The filing’s quarterly table, in millions of dollars, tells the story in one glance:

	Q3 2025	Q2 2026	Q3 2026
Revenue	$1,382	$11,555	$16,956
Operating income (loss)	—	+$974	−$567
Net loss	—	−$66,309	−$23,291

For the nine months ended September 30, 2026, revenue reached $33.24 billion, up roughly 1,200% from $2.55 billion in the same period of 2025. The nine-month net loss was $104 billion, and the nine-month operating loss was $1.26 billion. All quarterly figures for 2026 are unaudited.

Three things in that table deserve a closer look.

Growth is extraordinary, and it is slowing. Revenue grew about 47% from the second quarter to the third, down from about 144% growth the quarter before. Twelvefold year-on-year growth is still a pace few companies have ever matched — revenue was $1.38 billion in the third quarter of 2025 and $16.96 billion a year later — but the curve is bending. That is normal for a business of this size; it is also the first thing a careful reader should notice, because the valuation debate turns on how long the growth lasts.

The operating result flipped. In the second quarter, Anthropic made its first operating profit: $974 million. In the third quarter it posted a $567 million operating loss, and the operating margin fell from 8% to −3%. The swing is the whole story of the quarter, and it has a specific cause, covered below.

The adjusted numbers look better, but thinner. Excluding stock-based pay and certain charitable charges, Anthropic says it remained profitable on an adjusted basis: adjusted operating income of $2.46 billion in the third quarter, down from $2.24 billion in the second — but the adjusted margin narrowed from 19% to 15%. For the nine months, adjusted operating income was $4.08 billion, a 12% margin. Adjusted figures strip out real costs, so treat them as the company’s preferred lens rather than the full picture — but the direction of the margin matters whichever lens you use.

Why a $17 billion quarter still lost money

A company can sell $17 billion of services in three months and still lose money if the cost of delivering those services rises faster. That is what happened here. Total operating expenses rose to $17.5 billion in the third quarter, from $10.6 billion in the second.

The filing gives two specific reasons:

Short-term compute got expensive. Anthropic says it books most of its computing capacity 12 to 18 months ahead. When demand ran past that capacity, it bought about $5 billion of short-term compute in the quarter — capacity the filing says “currently costs significantly more” than capacity secured under multi-year deals. Compute and infrastructure expenses rose by $14.4 billion over the nine months, up 329% from a year earlier. The company is building its own dedicated infrastructure, called Anthropic Direct, which it says will let it replace costly short-term compute with cheaper capacity as it comes online — but that capacity is not online yet.

A one-off charge added to the hit. Anthropic cited about $2 billion in non-cash charges tied to an employee donation matching program. Together with the compute bill, that pushed the quarter from profit to loss.

There is a third pressure, quieter but important: prices are coming down. The filing says revenue growth from higher usage was partly offset by “a decrease of approximately $600 million from price reductions introduced as part of new model versions and contractual discounts” between the second and third quarters. Competition between model makers is starting to show up in the numbers. For Claude users, that is good news — model prices are still falling. For investors, it means growth has to come from selling more, not from charging more.

And then there is the net loss, the number most likely to be quoted out of context. The $23.29 billion net loss for the quarter — and the $104 billion nine-month net loss — comes mostly from changes in the fair value of convertible notes and derivative liabilities: an accounting effect tied to the rising value of those financing instruments, not cash spent running the business. In the third quarter alone, the filing records a fair-value loss of about $21.9 billion on those instruments. That is why the operating loss ($567 million) and the net loss ($23.29 billion) are so far apart, and why the operating line is the better guide to how the business itself performed.

The $895 billion question

The single largest number in the filing is not revenue or losses. It is the commitments: about $895.4 billion in purchase commitments and lease agreements, “substantially all of which is related to compute and infrastructure.”

Put that next to the cash pile. Anthropic held about $132 billion in cash and short-term investments at the end of September, and $146.6 billion in total liquidity once an available credit line is included. The commitments are roughly six times the cash on hand. That gap is the bet the whole listing rests on: that future revenue will be large enough, for long enough, to pay for computing capacity that has already been promised.

The timing of those commitments matters as much as the total:

• Roughly $70 billion comes due within a year.
• Nearly $450 billion comes due by the end of 2030.
• Purchase commitments alone jumped to about $314 billion, from $55 billion at the end of 2025.

The money flows to a small number of big suppliers. Broadcom tops the list at about $161 billion in equipment leases, followed by Google at about $121 billion, SpaceX’s AI unit at up to $120.5 billion and Amazon at about $110 billion. Google and Amazon are also major investors in Anthropic — which is why one analyst described the company, for the hyperscalers, as “at once a customer, competitor, sales channel and investment.”

More spending sits outside the $895 billion total. Anthropic also plans another $125 billion to $150 billion of Google chips and expects to spend more than $20 billion with AMD. Neither has a fixed dollar commitment, so neither is counted.

One more disclosure belongs in this section. Anthropic estimates that 5% to 7% of its revenue in the first nine months came from customers breaking its usage policy or commercial terms. The company stepped up enforcement in August and September and warned that the crackdown “may reduce our overall revenue and growth rates in future periods.” Cleaner revenue is easier to trust — but losing even 5% of a business running at $17 billion a quarter is a meaningful hole to fill while growth is already slowing. If you use Claude through a reseller or a high-volume setup, the same enforcement push is a reminder to check that your arrangement complies with the commercial agreement you signed — and if you run AI tools with access to your money, our AI agent money-safety checklist covers the spending-limit habits that matter most.

Who will actually control Anthropic

Anthropic is a public benefit corporation, and its governance is unusual for a company about to list. The filing keeps that structure in place, and buyers of the shares should understand what they are — and are not — getting.

• The Long-Term Benefit Trust will elect a majority of the board through a special class of stock.
• A separate Class F share, held by a company controlled by a co-founder, will carry 50.1% of the voting power on most stockholder matters.
• Buyers of the Class A shares — the ones that will trade as ANT — will therefore have a limited say over how the company is run.

That is not a flaw in the filing; it is a deliberate design, and Anthropic has been open about it from the start. But it is a material fact for anyone thinking of the listing as a normal purchase of a slice of a company. It is closer to a bet on a management team and a mission, with the usual shareholder levers largely removed.

Dilution is also real. Anthropic has 1.77 billion stock options outstanding and has set aside up to 2.95 billion shares for a new equity plan. New shareholders should expect their slice to shrink over time as those are issued.

The risk factor nobody expected

The most discussed passage in the amended filing is not a number. It sits in the Risk Factors section, and it was not in the original filing a week earlier.

Anthropic writes that its “increasing use of AI systems to accelerate our own research and development may cause the capabilities of our models to advance more rapidly than our ability to develop, validate, and implement corresponding safety, alignment, and security measures.” It adds that its internal estimates and expectations regarding future models, rather than current ones, “indicate that we may imminently meet our Responsible Scaling Policy’s definition of ‘dramatically accelerated progress attributable to AI.’”

In plain terms: the company believes its own AI may soon reach the threshold where AI sharply speeds up the building of AI — the point sometimes called recursive self-improvement. The same passage is careful to say that no current model has been designated as having crossed that threshold, and that the company is still assessing the data. The previous version of the filing said the company may not identify such an acceleration until after it has occurred; the amendment says it has in the past not identified one, and may in the future not identify one either.

The filing also reports, as covered in reporting on the document, that over 90% of contributions to Anthropic’s codebase in September 2026 were written by Claude itself. Whether that is a productivity triumph or a governance risk depends on your view — but it is the context in which the new risk language sits. For background on why Anthropic’s own researchers have warned that superintelligence may be hard to control, see our explainer on what AI alignment is.

What happens next

• Investor marketing: Anthropic is preparing to begin its roadshow — the formal pitch to institutional buyers — as soon as next week. Reports say the roadshow can start no earlier than November 10.
• Price range: a further amendment should fill in the blank price range and share count before the listing. Until then there is nothing to buy and no price to judge.
• Listing date: the filing sets no date. Reuters reported on November 3 that Anthropic is expected to go public before Thanksgiving, and Bloomberg reported that trading could begin as soon as November 16 — but both are reports, not confirmation, and the filing itself gives no date.
• Valuation: reports have pointed to a valuation of up to $2 trillion, while some current and prospective investors are now weighing about $1.5 trillion after the company shared financial details at a recent investor event. Treat both as reported, not settled.
• Access for US retail investors: the filing anticipates a retail selling group that includes Charles Schwab, Fidelity, Robinhood, Trade-PMR, SoFi and E*Trade. That would make this one of the more accessible large tech listings — but the price range is not set, so there is nothing to buy today.
• The fourth-quarter test: the next quarterly report will show whether new long-term capacity lowers compute costs or whether short-term buying keeps squeezing margins. That is the number to watch.

For readers following the listing, three things are worth doing now: watch the next amendment for the price range, which will show how far valuation expectations have shifted; track the 15% adjusted margin in future filings as the clearest sign of whether the long-term compute deals are paying off; and read the risk factors on enforcement and pricing before forming a view, since together they touched up to 7% of revenue and about $600 million in a single quarter.

Claude itself is the product most readers here actually use, and the filing is also a useful guide to where that product is heading: prices falling, capacity being built out, and a company that now has to answer to public shareholders while keeping an unusual governance structure. Our model guides — including the best AI models of 2026 and our Claude Haiku 5.5 review — cover what the models do; this filing covers what the company behind them costs to run.

Frequently asked questions

Sources

• Anthropic, PBC, Amendment No. 1 to Form S-1 (S-1/A), filed with the US Securities and Exchange Commission on November 2, 2026 — SEC EDGAR, accession no. 0001628280-26-068676, CIK 2130357
• Anthropic, PBC, Form S-1, filed October 26, 2026 — SEC EDGAR, accession no. 0001628280-26-067168
• AIF Intent — Anthropic’s updated IPO filing: $17B Q3 revenue, $567M loss and an $895B compute bill
• Deyron Labs — Anthropic files S-1 for Nasdaq IPO as ANT
• Reuters, November 3, 2026 (via reporting on the filing): listing expected before Thanksgiving
• Bloomberg, October 2026 (via reporting on the filing): marketing could begin the week of November 9, trading possibly from November 16

This article explains a public filing and is not investment advice. Figures are from Anthropic’s S-1/A as filed on November 2, 2026, and were checked on November 6, 2026. Share prices, dates and valuations described as reported have not been confirmed by the company.