Anthropic’s IPO Financials: $16.3bn Revenue, an $80bn Loss, and a $518bn Bill

The distinction matters before any number is discussed. According to reporting by Reuters, Anthropic filed a confidential draft Form S-1 with the SEC on 1 June 2026. A confidential filing is a real submission but a private one: the company and its regulator see it, and the public does not. The financial details circulating since late September 2026 come from journalists who have reviewed that draft, principally Reuters, the Wall Street Journal, the Financial Times, Bloomberg and AFP.

A public S-1 is expected imminently, and the listing has been reported for November 2026, after the 3 November US midterm elections. Anthropic has not confirmed the timing, the price, the share count or the size of the offering. Those three numbers are, as reported, left blank in the preliminary document and set during the roadshow — which is standard practice, not a sign of disarray.

So the correct frame for everything below is: these are figures reported from a draft prospectus by credible outlets, not audited statements the company has published itself. They may be revised before the public filing.

Quarterly 2026 revenue, as reported by Bloomberg and the Financial Times:

Read the last row against the first. Anthropic is reported to have earned slightly more in the first quarter of 2026 than in the whole of 2025. Year-over-year growth has been described in reporting as roughly a thirteenfold jump — a figure that is approximate and reported rather than an audited number, and which should be treated as a direction of travel rather than a precise multiple.

Three consecutive quarters of that shape is the single strongest fact in the filing. Revenue of this scale, growing at this rate, is not a pre-revenue story being sold on narrative. Whatever one concludes about the valuation, the demand is arriving.

For the first six months of 2026, the Wall Street Journal reported revenue of $16.3bn against a GAAP net loss of more than $80bn. For full-year 2025, the Financial Times, Reuters and AFP reported revenue of roughly $4.6bn, an operating loss of more than $8bn, and a net loss of roughly $42bn.

The 2025 pair is where the interesting question sits. An operating loss above $8bn and a net loss near $42bn means roughly $34bn of the damage happened below the operating line. Operating loss reflects the cost of running the business — training runs, inference, salaries, research. Items below that line are financing and non-operating in nature.

This is where analysis has to stop and label itself as analysis. The reporting establishes the gap; it does not establish its exact composition, and nothing here invents one. What can be said generally is that a gap of that shape in a late-stage private company is typically dominated by non-cash accounting items rather than cash going out of the door — the revaluation of preferred stock, derivative and warrant accounting, and share-based compensation are the usual candidates in companies that have raised at steeply rising valuations. A company whose paper value rose as fast as Anthropic’s did in 2025–26 can post enormous non-cash charges precisely because the valuation rose.

Why it matters for an investor: a $42bn net loss made mostly of non-cash marks is a very different object from $42bn of cash burn, and the two justify very different prices. The public S-1 will contain the cash flow statement that settles it. Until then, anyone treating the net loss as a cash figure is guessing, and so is anyone waving it away.

The number that deserves more attention than the loss is the one describing the future. Anthropic is reported to carry approximately $518bn in non-cancellable long-term commitments for cloud computing, compute and hardware. Amazon accounts for roughly $110bn of it, running from May 2026 to April 2036, with Google also among the counterparties.

Set that against the reported revenue. A company reporting $16.3bn of first-half revenue has contracted to spend something in the order of thirty times that figure, on a schedule reaching a decade into the future, for the single input its product cannot run without.

There are two honest readings, and they are not reconcilable by argument alone:

The confident reading. Compute is the binding constraint on frontier AI. Locking in a decade of it at pre-negotiated terms, before competitors bid the same capacity up, is exactly what a company that expects to need it should do. The commitment is evidence of conviction and of supplier relationships others cannot easily replicate.

The cautious reading. Non-cancellable is the operative word. If demand growth slows, if model efficiency improves faster than expected, or if a cheaper architecture arrives, the obligation does not shrink to match. It converts a variable cost into something much closer to fixed, and it does so for ten years — longer than the useful life of most of the hardware involved, and far longer than anyone can forecast this market.

The commitments also constrain strategy, not just cash. A company owing that much to specific cloud providers has materially less freedom to move workloads, renegotiate, or walk away than one buying compute at spot.

Nearly a quarter of 2025 revenue is reported to have come from just two customers, disclosed in the filing as a risk factor. Concentration of that degree is a standard disclosure item and a genuine vulnerability: the loss, renegotiation or in-housing of one relationship would be visible immediately in the top line.

It is also, in fairness, normal for enterprise businesses at an early stage of commercial scaling, and the 2026 quarterly curve suggests the base has broadened considerably since. The public S-1 should show whether concentration has fallen as revenue has multiplied. That single disclosure is worth more than most of the commentary that will surround the listing.

Anthropic has been reported as profitable on an adjusted basis and on course for a second consecutive quarter of adjusted operating profit. This is a meaningful operational milestone and it is not the same claim as profitability.

Adjusted figures exclude items management considers non-representative — most commonly share-based compensation, and often the non-cash charges discussed above. A company can be adjusted-profitable and GAAP-lossmaking by a very wide margin at the same time, and on the reported numbers Anthropic is exactly that.

Neither figure is dishonest. They answer different questions. GAAP asks what happened to the company’s equity; adjusted asks whether the core operation covers its own running costs. For a business whose main expense is building assets it will use for years, the gap between them is where the whole argument lives.

In May 2026, Anthropic’s Series H raised $65bn at a $965bn post-money valuation. The IPO is reported to target up to roughly $2 trillion, which would make it the largest listing ever attempted.

That is a doubling in roughly six months. Part of it is simply the revenue curve: Q3 revenue of $16.8bn against Q1’s $4.73bn is a materially different company from the one that priced in May. Part of it is the market the company is listing into, and part is the premium a liquid public market pays over a private round.

A useful discipline is to invert the question. At $2tn, what does the business have to become? Public-market software and platform companies have historically been valued across a wide band of forward revenue multiples, and a price of that order implies revenue growth continuing for years at a rate almost nothing sustains — and the $518bn of commitments converting into margin rather than consuming it. That is not an argument that the price is wrong. It is the specific thing a buyer at that price is underwriting.

Revenue has gone from roughly $4.6bn for all of 2025 to a reported $16.8bn in Q3 2026 alone. Adjusted operating profit has reportedly arrived earlier than almost anyone expected of a frontier lab. The compute commitments, on this reading, are a moat: they secure the scarce input for a decade at known terms while rivals bid against each other for what is left.

The losses are the cost of building an asset base, and the most severe of them appear to sit below the operating line, where non-cash items live. If the public S-1 shows cash burn far smaller than the headline net loss, the bear case loses much of its force in a single document.

And the demand picture is not speculative. Enterprises are paying, at scale, in quarters that keep getting bigger.

A $2tn valuation on a company with a reported first-half net loss above $80bn prices a decade of flawless execution into the opening trade. There is no margin for a slower year, a pricing war, or an architecture shift that makes today’s compute stack less valuable.

The $518bn of non-cancellable commitments is the hardest fact in the filing. It is a fixed obligation set against revenue that is, for all its growth, young and concentrated. Should growth decelerate, the obligation is still there, for ten years.

Customer concentration compounds it. And the company’s own prospectus — examined in detail in our companion piece on the risk-factor section — devotes roughly a third of its main body to risks, including the possibility that the product itself poses catastrophic harm. That is an unusual thing to underwrite at the largest valuation ever asked for.

Anthropic is not arriving alone. SpaceX filed its own S-1 in June 2026, and the period has seen a cluster of private giants moving to public markets at valuations that would have been implausible a few years ago.

This matters for pricing in both directions. A crowded calendar competes for the same institutional capital, which can discipline valuations; a hot one creates momentum that carries pricing higher than fundamentals alone would support. Which of those describes November 2026 will be visible in the aftermarket, not the prospectus.

For readers tracking how large investors position around events like this, our guide to reading 13F filings explains what institutional holdings data can and cannot tell you — and the Whale Tracker publishes the filings themselves. If you are weighing how to take a position at all, our practical guide to trading the ANT debut covers the reported timetable, retail access and the risks of each route.

The reported numbers describe a company growing faster than almost any at comparable scale, losing more money than it earns, and contractually committed to spending roughly thirty times its half-year revenue on compute over the next decade. All three statements are true simultaneously, and the valuation debate is really a debate about which of them dominates.

The single most useful thing a reader can do is wait for the public S-1 and go straight to the cash flow statement and the concentration disclosure. Those two pages will resolve more than every piece of commentary written before them, this one included.

Editorial Note: This article is educational and is not investment advice. It reports what has been disclosed in public reporting about a prospective listing and does not recommend any security. Anthropic has not confirmed IPO timing, pricing or share count; everything described as reported remains unconfirmed by the company.

Sources

• Reuters — review of Anthropic’s draft Form S-1 risk-factor section, reported 28 September 2026. reuters.com
• The Wall Street Journal — reporting on first-half 2026 revenue and net loss. wsj.com
• Financial Times — reporting on 2025 full-year results and quarterly 2026 revenue. ft.com
• Bloomberg — reporting on quarterly revenue and adjusted operating profit. bloomberg.com
• Agence France-Presse (AFP) — reporting on 2025 operating and net losses. afp.com
• U.S. Securities and Exchange Commission — Regulation S-K, Item 105 (Risk Factors). sec.gov