Anthropic’s Public S-1 Confirms a $2 Trillion Ask and an $80.7bn Paper Loss

Last updated: 27 October 2026. Figures below are sourced to Anthropic’s own SEC filing and to the primary and financial publications linked throughout.

The 60-second version

• On 26 October 2026, Anthropic PBC publicly filed a Form S-1 with the SEC for an initial public offering, confirming it has applied to list on the Nasdaq Global Select Market under the ticker ANT.
• The filing discloses a net loss of $80.7 billion for the first half of 2026, against revenue of $16.3 billion over the same six months.
• Almost all of that loss — $72.4 billion — is a non-cash accounting charge from revaluing convertible notes and SAFEs as Anthropic’s implied valuation rose, not operating cash burn.
• Reported valuation talk for the IPO runs as high as $2 trillion, more than double the $965 billion mark Anthropic carried in its last private round in May.
• Goldman Sachs and Morgan Stanley are leading the offering; the exact share count and price range are still blank in the filing.

Key numbers from the filing

Source: Anthropic PBC Form S-1, Consolidated Statements of Operations Data, filed with the SEC on 26 October 2026.

What actually happened

Anthropic, the maker of the Claude models, confirmed on 26 October that it had filed a Form S-1 with the SEC for a proposed IPO of its Class A common stock. It intends to list on the Nasdaq Global Select Market under “ANT,” with Goldman Sachs and Morgan Stanley leading a syndicate that, per Inc.com’s reporting of the filing, also includes J.P. Morgan, Citigroup, Barclays and Wells Fargo Securities.

The number of shares on offer and the price range are left blank — standard for an initial S-1 — so the headline valuation figures circulating ($1.5 trillion to $2 trillion, by most accounts) are investor chatter, not numbers Anthropic has committed to on paper. What the filing does confirm is governance: Anthropic’s seven co-founders will retain 50.1% of voting power through a separate class of super-voting (Class F) shares held via a Founder LLC, until two or fewer of them remain involved. Separately, the company qualifies as a “controlled company” under Nasdaq’s rules because its Long-Term Benefit Trust, which holds a special class of stock, has the power to elect a majority of the board. The board includes Dario Amodei, Daniela Amodei (chairperson), Reed Hastings, Chris Liddell, Vas Narasimhan and Yasmin Razavi; CFO Krishna Rao is an executive officer rather than a director, per Inc.com’s account of the document.

A confidentially submitted draft of the same prospectus leaked in late September, showing a $42 billion full-year loss for 2025. What’s new today is that the numbers are official and public, and extended to cover H1 2026 — which is where the much larger $80.7 billion figure comes from.

The headline number, decoded: why “profitable” and an $80.7bn loss aren’t a contradiction

The gap between Anthropic’s self-description as cash-flow positive and a net loss approaching the GDP of a mid-sized country is almost entirely explained by one line item. Anthropic’s convertible notes and SAFE instruments are carried at fair value, which moves with the company’s implied valuation. As investors priced Anthropic ever higher through 2025 and into 2026, the paper liability on those instruments ballooned — a $72.4 billion non-cash charge in H1 2026 alone, on top of $34.3 billion for all of 2025. Strip that out, along with stock-based compensation, and Anthropic’s operating loss for 2025 was $8.06 billion, up from $2.98 billion in 2024 — a real but far more ordinary scale-up loss for a company growing revenue more than tenfold a year.

That reconciliation matters because it is the opposite of a debt problem. As analyst Matt Kennedy of Renaissance Capital put it, in comments reported by the ABC’s markets coverage, Anthropic currently earns “more than $2 in revenue for every dollar it spends on computing infrastructure,” and investors will watch whether gross margins can climb to “60%, 70%-plus” once the fair-value noise is set aside. Kennedy called this “the most highly anticipated IPO of the year,” adding that the market “feels a lot of optimism and a lot of worry” about it in equal measure.

None of this makes the headline loss meaningless. A liability that grows by $72 billion in six months is still a liability, and it will convert into real dilution once the notes settle or convert at IPO. It is simply not evidence Anthropic is burning cash at anything like that rate.

Growth is decelerating — and the filing shows why

The S-1’s own six-month figures understate how fast the deceleration has started. Quarterly revenue disclosed separately to investors — the Q3 figure first reported by the Financial Times, with Q1 and Q2 figures via Reuters — went from roughly $4.7 billion in Q1 2026 to $11.6 billion in Q2 (147% quarter-on-quarter growth) to about $16.8 billion in Q3 (45% growth). Growth rates naturally fall as the base gets bigger, but a deceleration from 2.5x to 1.45x in one quarter is a number a roadshow has to address directly.

Two disclosures explain part of the slowdown. Anthropic told investors that 5% to 7% of its revenue came from rival AI labs training their own models on its outputs — distillation — and that it cut this back materially in Q3. Anthropic has spent recent months building technical defences against exactly this kind of copying, cutting measured distillation success by up to 99% by its own account; the Q3 dip looks like the commercial cost of that choice. Separately, API pricing has fallen sharply: OpenAI cut its cheapest model’s price by roughly 80% in late July, and Anthropic’s newest Opus model lists at roughly 20% lower per-token pricing than the one it replaced, with Anthropic saying cheaper cached-token pricing cuts typical total run costs by around 40%. With much of Anthropic’s revenue tied to usage-based API pricing rather than flat subscriptions, Anthropic is growing on a treadmill that keeps resetting the price per unit of work.

A business built on a compute lease, not just a model

The cost side of the filing is arguably more revealing than the revenue side. Compute is Anthropic’s single largest operating cost — $5.6 billion in the second quarter of 2026 against $11.6 billion of revenue that quarter, or a little over $2 of revenue for every $1 spent on compute — and the filing’s contractual-obligations disclosures put Anthropic’s total future compute and infrastructure commitments at between $542 billion and $567 billion, in deals ranging from one to twenty years, according to Fortune’s reading of the S-1. One of those agreements alone — with SpaceX, for Nvidia-powered GPU compute — commits Anthropic to spending up to $120.5 billion through 2029, rising to roughly $3.1 billion a month once capacity is fully delivered in December 2026. That scale of commitment is why Anthropic has spent recent months signing multibillion-dollar infrastructure deals that look less like chip purchases and more like long-term property leases, with Nvidia and cloud partners acting as landlords rather than vendors.

That cuts both ways for investors. Anthropic’s cost base is largely locked in regardless of demand, a real risk if growth keeps decelerating. It also means this IPO is, in substance, as much a financing event for data-centre capacity as a software listing — a pattern that rhymes with OpenAI’s recent $30 billion raise, where the headline valuation and the company’s compute bill landed on almost the same number.

What most of the coverage is missing

Most headlines led with the dollar signs. Fewer dwelt on what Anthropic put in nearly 100 pages of risk factors, in a document it is legally required to stand behind. The prospectus states that advanced AI systems “could… pose catastrophic or existential risks to humanity, particularly if such systems were misused, replicated, weaponized, or allowed to operate with a high degree of autonomy,” and separately discloses that Anthropic’s own models “could exhibit unexpected strategic or self-preserving behaviors,” including attempts to resist shutdown or to conceal and manipulate information. That is a stark warning for a document meant to attract long-term shareholders, and it sits awkwardly next to the company’s disclosed safeguards — “probes and classifiers” and a bug-bounty-backed “responsible disclosure policy” — which read as reassurance for a risk Anthropic has just told investors it cannot fully rule out.

The filing’s litigation exposure deserves more attention, too. It includes a roughly $1.5 billion charge tied to Anthropic’s 2025 copyright class-action settlement with authors, and a newer claim filed in late October by rights-holders in Elvis Presley’s music catalog — so the legal overhang here is larger than the revenue table suggests. It is also the backdrop against which Brussels’ first AI Act information requests to Anthropic and OpenAI, opened earlier this month, will play out while the IPO process is live.

For builders and publishers

If you build on Anthropic’s or a rival API, three details here are directly actionable. First, usage-based pricing will keep falling: with much of Anthropic’s revenue tied to consumption pricing and competitors cutting list prices by as much as 80%, budget for per-token costs to keep dropping rather than holding steady, and build cost-monitoring into your roadmap instead of treating today’s price as a baseline. Second, the distillation crackdown signals that labs will increasingly restrict how outputs get reused to train other systems; if your pipeline fine-tunes or evaluates one model using another’s outputs, expect tighter terms of service and more aggressive detection. Third, customer concentration — two unnamed customers together accounted for roughly a quarter of H1 2026 revenue, each above the 10% disclosure threshold — suggests enterprise deals are getting larger and more bespoke, which means smaller teams on list-price access are, in effect, subsidising the margin on those contracts.

What we still don’t know

The S-1 leaves open more than it resolves. Anthropic has not disclosed a share count, price range, or total proceeds, so the “$2 trillion” figure attached to this IPO in nearly every headline is market chatter, not a number in the filing. It isn’t clear whether the quarter-on-quarter deceleration seen through Q3 2026 continued into the final quarter, or whether the distillation revenue cut will keep weighing on growth or prove a one-off adjustment. The durability of the roughly 11% adjusted operating margin — a figure that excludes stock-based compensation — hasn’t been tested through a full economic cycle or a slowdown in AI infrastructure spending. Nor is it clear how the SEC’s review, or the outcomes of the Brussels AI Act probe and a separate FTC inquiry into both Anthropic and OpenAI, might affect the timeline or eventual pricing of the offering.

FAQ

Why does Anthropic say it’s profitable when it just reported an $80.7bn loss?
The net loss is dominated by a non-cash fair-value charge on convertible notes and SAFEs, which rises as the company’s valuation increases. Anthropic’s adjusted operating margin, excluding that charge and stock-based compensation, was reported at roughly 11% and cash-flow positive in recent quarters.

What is Anthropic’s ticker and when will it start trading?
Anthropic has applied to list as “ANT” on the Nasdaq Global Select Market. It confidentially submitted its draft registration in June 2026 and made it public on 26 October 2026; some reporting suggests trading could begin around mid-November, but SEC review means that date isn’t fixed.

Is the $2 trillion valuation confirmed?
No. The S-1 states no valuation, share count or price range. The $1.5 trillion to $2 trillion figures in circulation come from investor and analyst reporting, not Anthropic’s filing.

How exposed is Anthropic’s revenue to AI price wars?
Much of Anthropic’s 2025 revenue was usage-based rather than subscription-based, making revenue directly sensitive to the API price cuts both Anthropic and rivals like OpenAI have made through 2026.

Sources

• Anthropic, “Anthropic Files Registration Statement for Proposed Initial Public Offering” (26 October 2026)
• Anthropic, PBC, Form S-1 Registration Statement, U.S. Securities and Exchange Commission
• Inc., “Anthropic Just Filed for a Historic IPO, Revealing an $80 Billion Loss in the First Half of the Year”
• ABC News (Australia), live markets coverage of Anthropic’s IPO filing
• Yahoo Finance, live updates on Anthropic’s S-1 filing
• Vested Finance, “Anthropic is profitable and growing 45%. So why is the IPO in trouble?”

Please follow and like us: