Anthropic IPO Filing: Revenue, Losses and Who Controls It

What does Anthropic's IPO filing show?

The filing shows a company with very fast revenue growth, large and rising costs, huge long-term commitments for computing power and an unusual control structure that keeps decisions with the founders and a trust ^[1][5]. Anthropic, the maker of the Claude models, filed its registration statement (the S-1, the document a company must publish with the US Securities and Exchange Commission before it can sell shares to the public) in October and amended it on 10 November 2026 ^[1][5].

The amended filing added the numbers most readers search for: the price range, the number of shares and the results for the first nine months of the year ^[1][4][5]. Anthropic says it will sell 801,351,354 new shares, existing holders will sell 23,648,646 more, and the underwriters may buy up to 123,750,000 extra shares within 30 days ^[2].

The basic timetable, price range and buying routes are covered in the Pouyam guide Anthropic IPO: Date, Price Range and How to Buy ANT Shares. This guide goes one step further and reads the parts of the filing that decide what a share actually gives its owner: revenue, losses, spending promises, voting rights and when insiders may sell.

Everything below is as of 15 November 2026. The prospectus is marked "subject to completion", which means figures such as the price and the number of shares can still change before pricing ^[1].

How much revenue does Anthropic make?

Anthropic reported $33.2 billion of revenue for the nine months to 30 September 2026, compared with $4.6 billion for the whole of 2025 ^[5]. The first version of the filing, which covered only the six months to 30 June, listed $16.3 billion, so roughly half of the nine-month total came in the third quarter alone ^[5].

That growth is the main reason investors are interested. Anthropic sells access to its Claude models to businesses and developers through its own products and through cloud partners, and to individuals through paid plans. The filing describes the ability to keep building the most capable models as central to the business and lists failure to do so as a key risk ^[11].

Revenue is not profit. Training and running large models needs enormous amounts of specialised chips and electricity, and those costs rise with demand. The next section shows how much of the revenue is absorbed by those costs.

Two cautions when reading the revenue figure. First, it is a nine-month number for a company growing quickly, so comparing it with a full calendar year flatters the change. Second, the figure comes from a preliminary prospectus; the final prospectus filed after pricing is the version that counts ^[1].

Is Anthropic profitable?

No: Anthropic recorded an operating loss of $1.3 billion for the first nine months of 2026, smaller than the $5.2 billion loss in the same period of 2025, and it lost $536 million in the September quarter alone ^[5]. In the June quarter it had reported operating income of $974 million, so the third quarter reversed that gain ^[5].

According to the filing, as reported by Yahoo Finance, the swing came mainly from a $5 billion charge for computing capacity bought under short-term commitments and $2 billion of non-cash charges linked to the company's donation-matching programme ^[5]. Non-cash means no money left the company at the time, but the charge still reduces reported results.

An operating loss is the result before interest, taxes and some financial items. It is the clearest measure of whether the core business covers its own costs. For Anthropic, one profitable quarter followed by a loss-making one shows how sensitive results are to the timing of large compute purchases.

For readers comparing AI companies, the useful question is not whether a firm is profitable today but whether costs grow more slowly than revenue over several quarters. The filing gives only three quarters of 2026 data, so that trend cannot yet be judged with confidence.

What are Anthropic's $895 billion commitments?

Anthropic disclosed aggregate committed spending of $895.4 billion, which it says relates mainly to compute and infrastructure ^[5]. Of that total, $447.5 billion falls due before 31 December 2030, the rest between 2031 and 2045, and $69.8 billion is due within the next 12 months ^[5].

These are contracts for future chips, data-centre capacity and cloud services, signed so that the company can train and serve its models years ahead. They are not debts in the usual sense, but they are obligations the company must fund whether or not revenue keeps growing.

The size of these promises explains several other parts of the deal. Alongside the IPO, Anthropic agreed to sell $26 billion of non-voting Class N stock at the IPO price to AMD, Broadcom, Google, Micron, Microsoft, Nvidia and TSMC, with Google and Nvidia investing $10 billion each ^[4]. Most of these buyers are also suppliers of the computing power the company needs ^[10].

TechCrunch reported that Anthropic estimates its total net proceeds from the IPO and the placement at between $84.6 billion and $93.7 billion, the higher figure if the underwriters use their option to buy extra shares ^[4]. Even that amount is only about a tenth of the $895.4 billion commitment total, and running costs come on top, which is why investors will watch cash flow closely after listing.

For background on why computing power has become the limiting factor for AI companies, see the Pouyam guide AI Data Centers and Power: Why Electricity Is the Limit.

Who controls Anthropic after the IPO?

The seven co-founders and an independent trust keep control: a single Class F share held by the founders' company carries 50.1% of the votes on general matters, and the Long-Term Benefit Trust elects directors who form a majority of the board ^[1][11]. Buyers of Class A shares receive one vote per share, but together they can never reach a majority of general votes ^[8].

The filing names the co-founders as Dario Amodei, Daniela Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish and Christopher Olah ^[11]. Morningstar explains that each founder has one vote inside the founders' company and a simple majority decides how the share is cast, so four of the seven can determine the outcome of an ordinary shareholder vote ^[8].

The Long-Term Benefit Trust is described in the filing as a body of five financially disinterested trustees with expertise in AI safety, national security, public policy and social enterprise ^[11]. Through Class T stock it elects and removes the Class T directors ^[1]. Morningstar describes this as four of seven board seats ^[9].

Anthropic is also a public benefit corporation under Delaware law, which requires the board to balance shareholders' financial interests with the interests of people affected by its conduct and its stated purpose: the responsible development of advanced AI for the long-term benefit of humanity ^[1]. In practice, the board is not obliged to choose the most profitable option in every case ^[9].

The company will be a "controlled company" under exchange rules, according to a page-by-page review of the filing, which allows it to skip some board-independence requirements, although the review says the company does not expect to rely on that exemption except for its nominating and governance committee ^[12].

How are the different Anthropic share classes different?

Only Class A and Class N shares carry economic rights such as a share of profits; the other classes exist mainly to allocate control ^[1][8]. The table below summarises each class as described in the amended filing.

Class N shares are held mainly by strategic business partners and generally convert to Class A when sold to an unrelated buyer ^[1]. Morningstar notes that such conversions add voting Class A shares and can dilute the voting weight of existing Class A holders, although the founders' 50.1% stays fixed ^[8].

Morningstar also points out that a separate Public Benefit Fund already holds 5% of the company and can receive up to 2% more each time the share price rises 50%, which would dilute other holders after large gains ^[9]. The filing warns that the multi-class structure may keep the stock out of some stock-market indexes, which matters because index funds are large buyers ^[9].

When can Anthropic insiders sell their shares?

Most insiders are locked up for 180 days and the co-founders for up to 365 days (or until an earlier results date), but the filing allows staged early releases, including up to 10% of locked shares from the first trading day for some holders ^[12]. A lock-up is an agreement not to sell shares for a set period after an IPO, so that a flood of insider sales does not hit the price straight away.

According to a review of the prospectus, non-founder directors, officers and some staff may sell up to 10% from the first trading day and a further 35% after the 2026 annual results; the co-founders get no release until the first-quarter 2027 results or day 180, and then up to 29% ^[12]. The lead banks, Goldman Sachs and Morgan Stanley, can release shares early ^[4][12].

The founders are already selling a small part of their holdings in the IPO itself: TechCrunch counted nearly 3.4 million shares each, worth about $250 million each at the $74 midpoint of the range ^[4].

Lock-up dates are worth noting in a calendar because the end of a lock-up can bring extra selling. The filing itself lists the anticipation of lock-up releases among risks to the share price, as the same review notes ^[12].

What happens next with the Anthropic IPO?

The roadshow, the period when the company and its banks present the deal to investors, runs until Wednesday 18 November 2026, when the final price is expected to be set; trading on Nasdaq under ANT is expected to begin on Thursday 19 November ^[2][5][7]. The final price can land inside, above or below the $70 to $78 range ^[1].

Demand signals so far are reports, not filings. Bloomberg reported on 12 November that orders were about six times the shares available, and Reuters reported on 13 November that interest was outpacing the early days of SpaceX's roadshow, while noting that details could change ^[6][7]. Bankers told investors that existing backers would be prioritised, so some buyers may receive fewer shares than they asked for ^[6].

Unlike SpaceX, which set a fixed price of $135 during marketing, Anthropic is using a conventional price range, which gives large investors more influence over the final price ^[10]. Bloomberg also noted a weaker market backdrop: the Nasdaq 100 was down about 7% over three months as of 14 November ^[10].

After pricing, the company files a final prospectus with the exact price and share count; that document, not news reports, is the reference for any figure. This guide explains a public filing and is not investment advice; anyone considering buying should read the prospectus and consider speaking to a licensed adviser.