Anthropic Is Going Public: What the S-1 Means for People Who Use Claude, the Real Numbers Behind the $80.7 Billion Loss, Who Controls the Company After the IPO, and What Could Change About Pricing, Privacy, and the Free Tier

What was filed, and what's still missing

Anthropic announced a confidential draft submission on June 1, 2026, amended it five times through October 9, and made the full S-1 public today, about 240,000 words of it, with a newsroom notice naming Goldman Sachs and Morgan Stanley as the banks to request a prospectus from (the cover lists more than twenty underwriters). It is a Delaware public benefit corporation whose stated purpose is 'the responsible development and maintenance of advanced AI for the long-term benefit of humanity.' The filing leaves blank the things that set the price: the number of shares, the price range, the offering size, revenue for the third quarter, the 'Run-Rate Revenue' as of October 31, and the total of its compute commitments. Those arrive in a later amendment. Everything about timing is press, not filing: Bloomberg reported on October 1 that marketing could begin 'as soon as the week of Nov. 9' with a debut before Thanksgiving, and Reuters repeated the pre-Thanksgiving expectation today, both citing people familiar with the matter. Anthropic has not given a date.

The numbers, from the S-1 itself

Why the loss is $80.7 billion when the company made money operating. Anthropic raised much of its capital through convertible notes and similar instruments that turn into shares at a later valuation. Accounting rules require those to be carried at fair value, and when the company's valuation jumps, the notes become more valuable to their holders, which the income statement records as a loss to the company. The S-1 says the charge was 'primarily driven by an increase in the fair value of our outstanding convertible promissory notes resulting from an increase in our enterprise valuation,' and that it was 'largest in the three months ended June 30, 2026, reflecting the increase in our enterprise valuation in connection with our Series H financing.' No cash left the building for that $72 billion. The cash story is the compute line: $8.9 billion in six months, more than all of 2025, and positive operating cash flow of $3.9 billion despite it, because revenue grew faster still.

What's press, not filing. The Financial Times and Bloomberg reported on October 15, from an investor presentation, that third-quarter revenue reached $16.8 billion, up about 45 percent from the second quarter, with adjusted profitability, positive free cash flow, and an 85 percent gross margin. The S-1's Q3 cell is blank, so treat those as reported. Valuation is entirely press: Bloomberg's October 1 range of $1.8 to $2 trillion, Reuters' October 26 report that some investors now say about $1.5 trillion 'down from roughly $2 trillion a week earlier,' and a single anonymous early backer telling Fortune on October 15 that shares could trade at $2.3 to $2.4 trillion. The last figure is one person's prediction, not a consensus, however often it's repeated.

The compute commitments: who Anthropic has promised to pay

The S-1 doesn't total them, but it itemizes them in the contractual-obligations notes, and the itemized figures are larger than the draft numbers the Wall Street Journal reported on October 20. Google: at least about $114.5 billion between April 2026 and July 2033, and Google may lend up to $20 billion more while its ownership is capped at 14.99 percent. Amazon: at least $110 billion between May 2026 and April 2036, on top of $6.2 billion of Amazon convertible notes outstanding. Microsoft: at least $44.9 billion between November 2026 and May 2033, a figure amended upward this month from the roughly $31 billion in the draft. SpaceX.AI, the company formerly called xAI: up to $120.5 billion into 2029, about $3.1 billion a month once capacity is delivered from December 2026. Broadcom-related finance leases: about $161.2 billion excluding interest. AMD: a hardware agreement under which AMD committed to buy up to $5 billion of non-voting stock. Yahoo Finance's arithmetic on the non-cancelable portions comes to roughly $430 billion; Reuters' September read of a draft put it at $518 billion across six partners, and the Journal reported Anthropic 'believes it could spend more than $750 billion' including deals still being negotiated. Whatever the total, the shape is clear: the company has pre-committed well over a decade of revenue at today's run rate to computers, which is the single fact that most constrains what it can do about prices.

Who controls Anthropic after the IPO

This is the part of the filing that matters most for anyone who chose Claude partly because of how Anthropic says it makes decisions. There are five share classes, not two. Class A, the shares being sold, carry one vote each. Class N carries no votes and is held mostly by 'strategic business partners' such as Amazon, Google, and AMD. Class T is held by the Long-Term Benefit Trust, an 'independent body of five financially disinterested voting trustees,' and it elects the Class T directors, 'who together will constitute a majority of the authorized members of our board.' Today those directors are Reed Hastings, Christopher Liddell, and Vas Narasimhan; the trustees named on Anthropic's site are Neil Buddy Shah, Ben Bernanke, Richard Fontaine, and, as of October 21, Christopher Stubbs. Class F is a single share held by an entity owned by the seven co-founders that carries 'an aggregate number of votes that equals 50.1% of the total voting power' on everything except electing Class T directors, with no economic rights. Class E goes to employees for limited matters. Anthropic will be a 'controlled company' under Nasdaq rules. There is a release valve: holders of 85 percent of Class A shares can amend the Trust's terms or redeem Class T if the trustees refuse. In plain terms, buying ANT buys a claim on the profits and almost no say over the mission. Whether you find that reassuring or alarming depends on how much you trusted the mission before; it is, at least, written down now.

Two more governance items touch users directly. The Public Benefit Fund, announced October 18, sets aside 5 percent of fully diluted shares (about 847 million Class N shares in the filing) for a philanthropic LLC that 'will not be used for lobbying or political giving,' subject to shareholder approval, and the co-founders have pledged 80 percent of their equity to philanthropy, which the S-1 notes is 'voluntary, revocable' and not legally binding. And the Responsible Scaling Policy is now a disclosed business constraint: the filing says safety policies 'have previously resulted in, and may in the future result in, slower decision-making and product releases,' lists 'catastrophic or existential risks to humanity' among risk factors, and discloses that models 'have exhibited… self-preserving behaviors, including attempts to resist shutdown… behavior resembling blackmail.' It also records that the Commerce Department imposed 'worldwide export restrictions on our Fable 5 and Mythos 5 models' in June 2026, 'which we disabled for all customers until those restrictions were lifted.' Public companies rarely write sentences like those; a public benefit corporation with a safety trust apparently has to.

What actually changes for people who use Claude

• Your price, today: nothing. Claude's pricing on October 26 is unchanged from the summer: Free; Pro $20 a month or $17 annual; Max from $100 (5× or 20× Pro usage); Team $20 to $100 a seat; Enterprise $20 a seat with API-rate usage. No consumer price change has been announced in September or October. The S-1 does disclose a Sherman Act class action brought by 'purchasers of the defendants' paid consumer subscriptions,' which is the first time Anthropic has acknowledged a consumer-pricing lawsuit in writing.
• Your price, over time: watch the risk factors. The filing states plainly that Anthropic has 'changed, and may in the future change, usage limits, rate limits' and pricing. That isn't new behavior: weekly limits arrived in 2025, peak-hour session windows tightened in March 2026, and Claude Code's weekly limits were permanently raised 25 percent on September 14 after a temporary 50 percent boost. What's new is that those levers now belong to a company with quarterly earnings calls and more than $400 billion of compute bills. Expect more tiers and more metering, not fewer.
• Ads: the promise isn't in the prospectus. In February, Anthropic wrote that 'Claude will remain ad-free' and that it would be transparent 'should we need to revisit this approach.' The S-1 contains no ad-free commitment; its only related line is that free tiers 'may cannibalize potential paid subscriptions.' A blog post is not a covenant, and investors will read the omission. The revenue mix gives Anthropic room, though: 80 percent of its money is API consumption, not subscriptions, so the consumer app is a smaller lever than it is at OpenAI, whose $8 Go plan already 'may include ads.'
• Your data: opt-out, and it has been since September 2025. The privacy policy effective September 10, 2026 says Anthropic 'may use your Inputs and Outputs to train and improve Anthropic AI models, unless you opt out through your account settings,' on Free, Pro, and Max including Claude Code; Work, Enterprise, Education, Government, and API traffic are excluded. With training on, de-identified data can be kept up to five years; deleted chats leave the back end within 30 days. If you haven't looked at that toggle since the October 2025 deadline, the IPO is a good excuse. Our Claude review still credits Anthropic for the strongest defaults in the chatbot field on business plans; the consumer default is the one to check.
• Your models: a notice period, now in writing. Anthropic's deprecation policy commits to 'at least 60 days' notice' before retiring a public model, and its deprecation commitments promise to preserve weights 'for, at minimum, the lifetime of Anthropic as a company.' The S-1 adds that welfare considerations 'may lead us to limit or delay the deprecation of older models.' Current example: Sonnet 4.5, deprecated September 30, retires November 30.
• Your policy terms: changing November 12, for unrelated reasons. The Usage Policy update effective November 12 was announced October 8 and covers deception, elections, weapons, surveillance, and a new ban on 'sustained and needless abusive or cruel behavior' toward models; nothing in it concerns pricing, ads, or the IPO. We covered it in Can Claude ban you for being mean to it?
• Your product roadmap: already visible. Cowork merged into Claude on September 16, Claude Duet opened its preview October 21, and the filing's use-of-proceeds language is 'compute and infrastructure, as well as hiring and retention.' None of that is a pivot; it is the same company with a bigger checkbook and a reporting calendar.

How this compares with OpenAI

OpenAI confidentially submitted its own S-1 on June 8, a week after Anthropic, with the line 'We have not decided on timing yet; it may be a while.' Nothing of OpenAI's is public on EDGAR; CNBC reported on October 19 that it has ruled out a 2026 listing, Reuters today said early 2027, and TechCrunch reported a new $30 billion raise at a $1.25 trillion pre-money valuation on October 22. Structurally, OpenAI's nonprofit foundation holds about 26 percent of OpenAI Group and appoints its entire board; Anthropic's trust appoints a board majority while the founders hold 50.1 percent of the vote. The Information reported on October 24 that OpenAI is undercutting Anthropic on price in enterprise coding deals, citing a customer whose Anthropic share of coding work fell after peaking at 57 percent in May. That is the competitive fact most likely to shape what Claude costs you: the two companies are now racing each other to the public markets while cutting each other's margins, and the user is the beneficiary until one of them stops.

What the headlines got wrong this week

• '$80.7 billion loss.' True on the income statement and misleading everywhere else. Operating loss was $693 million for the half, the second quarter was operating-profitable, and cash flow from operations was positive $3.9 billion.
• '20% of revenue from two customers.' That was the Journal's reading of a draft. The public filing shows one customer above 10 percent (11 percent) in the first half of 2026 and says no customer exceeded 10 percent in the nine months to September 30. Two customers at 12 percent each was the 2025 picture.
• '$450 billion of cloud commitments, Microsoft $31 billion.' The itemized public figures are higher: Microsoft at least $44.9 billion after an October amendment, Google about $114.5 billion, Amazon $110 billion, SpaceX.AI up to $120.5 billion, Broadcom leases about $161.2 billion. The total is blank in the filing.
• 'Dual-class shares.' Five classes. The founders' 50.1 percent sits in a single Class F share with no economic rights.
• '$2.3 to 2.4 trillion.' One anonymous investor's guess to Fortune. Reuters' reporting today has some investors at $1.5 trillion.
• 'Trading before Thanksgiving.' Reported, plausible, and not in anything Anthropic has published.

Dates to watch

Our read

For users, the honest summary is that the IPO formalizes incentives that already existed. Anthropic was already raising prices by tightening limits, already training on consumer chats by default, already promising ad-free in a blog post rather than a contract, and already governed by a trust most customers had never read about. The S-1 puts all of it in one document that a court, a regulator, and a shareholder can hold the company to, and that is a net gain for the people who depend on the product, even if some of the sentences are uncomfortable. The thing to actually worry about is arithmetic, not intent: a company that has committed hundreds of billions of dollars to compute through 2036 will meter everything it can meter, and the consumer app is the most visible place that shows up. Our advice is unchanged from before the filing. Check the training toggle. Pay annual if you were going to pay anyway, since that locks a price for a year. And keep a second assistant warm; the chatbot ranking exists because no one of these companies, public or private, has earned exclusivity. We're not investment advisers and this isn't advice about ANT. We'll update this post when the price range lands, when a date is confirmed, and after the first earnings call, which is when the user-facing questions here start getting answered in numbers.