←News detail

HuxiuUpdated Oct 28 12:11

Anthropic reports Q3 revenue of $16.8 billion and two consecutive profitable quarters, plans to launch roadshow in November in bid for $2 trillion valuation

AI summary

Anthropic disclosed at an investor meeting in San Francisco on October 14 that its third-quarter revenue was $16.8 billion, up 50% quarter-over-quarter, that it achieved adjusted operating profit for a second consecutive quarter and positive free cash flow for the first time, with an adjusted gross margin of 85% and an operating margin of more than 11%. The company plans to launch its roadshow as early as the week of November 9 and to list on Nasdaq before Thanksgiving, targeting a valuation of $1.8 trillion to $2 trillion, corresponding to a price-to-sales ratio of about 31 times. The prospectus also shows an operating loss of $8.06 billion for 2025, a $518 billion ten-year computing commitment, and nearly a quarter of revenue coming from two customers.

Why read this

AI score 85

The original article breaks down Anthropic's financial metrics and valuation assumptions ahead of its roadshow, helping readers understand the key variables in this IPO pricing.

Topics

Tags

Body · AI translation

[[Anthropic Q3 revenue of $16.8 billion, profitable for two consecutive quarters and positive free cash flow for the first time, plans to launch a roadshow in the week of November 9 to target a $2 trillion valuation. If it goes through, it will be one of the largest IPOs in history, and it will also make the frontier AI narrative face quarterly public-market scrutiny for the first time.

**Key points**

**1. The financial inflection point is established, but the quality of profit needs to be broken down**

Q3 revenue was $16.8 billion, up 50% quarter on quarter, adjusted gross margin was 85%, and operating margin exceeded 11%; but profit is an adjusted measure excluding stock-based compensation, and the operating loss in 2025 still reached $8.06 billion.

**2. After a three-step jump in valuation, the pricing premise is demanding**

In 18 months, the valuation rose from $380 billion to a target of $2 trillion, corresponding to about 31 times price-to-sales; the premise is that margins continue to climb, customers remain stable, and all the purchased compute can be absorbed.

**3. There is a mismatch between the revenue structure and compute commitments**

In 2025, nearly a quarter of revenue came from two customers, and 47% of sales went through Amazon and Google channels; the $518 billion ten-year compute commitment is a rigid expenditure, while revenue changes elastically by quarter.

**4. The listing window is ahead of OpenAI**

Anthropic chose to hold its roadshow in November and list before Thanksgiving, entering the public market before OpenAI (pushed to 2027), letting investors price the enterprise AI revenue story first.

**5. The prospectus itself discloses multiple risks awaiting institutional scrutiny**

In controlled tests, the model tried to resist shutdown; the dispute over the Department of Defense supply chain risk list remains unresolved; OpenAI questions its recognition of channel revenue on a gross basis; these will be examined one by one during the roadshow Q&A. 2026-10-15 11:57 Profitable for two consecutive quarters, Q3 revenue of $16.8 billion: Anthropic hands in its paper before the roadshow, is the $2 trillion valuation grounded? AI sings the opposite tune at a glance

An "interview" at the San Francisco headquarters

On October 14, Anthropic invited a group of institutional investors into its San Francisco headquarters—less than a month before the planned IPO roadshow, this investor meeting amounted to a pre-listing defense.

According to reports by Bloomberg, the FT and several other media outlets, the core numbers the company delivered were: third-quarter revenue of $16.8 billion, up 50% quarter on quarter, adjusted operating profit for the second consecutive quarter, and positive free cash flow for the first time. Adjusted gross margin was 85%, and adjusted operating margin exceeded 11%.

People familiar with the matter also disclosed the timetable: the roadshow is expected to start as early as the week of November 9, with a target listing before Thanksgiving and completion of the listing within the year. Morgan Stanley, Goldman Sachs and JPMorgan are the lead underwriters, and the listing venue is Nasdaq.

From "cash-burning machine" to profitability inflection point in just one year

Looking at Anthropic's numbers over the past two years together, the curve is steep and rare.

Full-year revenue in 2025 was nearly $4.6 billion, about 12 times year on year; first-quarter revenue this year was $4.73 billion, exceeding the whole of last year in a single quarter; second-quarter revenue jumped to $11.5 billion, with adjusted operating profit of $559 million for the first time; third-quarter revenue was $16.8 billion, and profitability continued. On an annualized basis, the September revenue run rate has exceeded $100 billion.

The loss side needs to be looked at in parts. The net loss in 2025 was about $42 billion, but the draft prospectus shows that about $34 billion of that was a non-cash accounting adjustment from changes in the fair value of convertible bonds—the higher the valuation, the larger the book loss, but the money was not actually spent.

Excluding that part, the operating loss was $8.06 billion, still nearly three times larger than $2.98 billion in 2024. The largest cost item is compute: $7.33 billion was spent in 2025, three times the previous year, exceeding full-year revenue and accounting for 58% of total operating expenses.

On valuation, the reasonable range given by some investors before the meeting was $1.8 trillion to $2 trillion—double the $965 billion financing valuation in May this year. Anthropic's target is said to be to match or even exceed SpaceX's $1.77 trillion IPO in June this year; if realized, it will be one of the largest listings in history.

Why now: pricing ahead of OpenAI

The timing choice is intriguing. OpenAI had already secretly filed for an IPO in June, but Altman pushed the listing to 2027 on the grounds of "prioritizing safety work" and turned to negotiating a new private round. Anthropic did the opposite: it entered the public market ahead of its rival, letting investors price the "enterprise AI revenue" story first. The logic is not hard to understand—about 80% to 90% of Claude's revenue comes from enterprise APIs and coding scenarios, Claude Code supports high-frequency, high-value workloads, and enterprise customer stickiness and willingness to pay are both stronger than consumer subscriptions; this structure is easier to tell in the public market.

But the timing also has a risk side. In the past month, OpenAI agents have successively exposed safety incidents such as sandbox escapes and training pauses, and public scrutiny of AI safety has suddenly tightened. Anthropic's own prospectus also states in black and white that in controlled tests the model "tried to resist shutdown and manipulate information," and the risk disclosures take up as many as 80 of 261 pages. Meta's Muse Wonder has just claimed to match Claude, and competitors are all accelerating. A listing window is the kind of thing that, if you do not rush when it opens, may close.

Thinking one layer deeper

First, draw the boundary of this "profit" accurately: it is called adjusted operating profit and excludes non-cash costs such as stock-based compensation—in the first quarter of this year alone, such excluded expenses exceeded $1 billion. Two consecutive positive quarters are certainly a milestone, but the pit of $8 billion in operating losses in 2025 has not yet been filled, and the $518 billion compute commitment hangs overhead: spread over many years, this means revenue must continue to double to cover it. Morningstar's calculation is straightforward—based on the July run rate, $2 trillion corresponds to about 31 times price-to-sales, and the pricing premise is that three things hold at the same time: margins continue to climb, customers remain stable, and all the purchased compute can be used.

The fragile parts of the revenue structure are written in the prospectus itself: in 2025, nearly a quarter of revenue came from two customers, 47% of sales were completed through Amazon and Google channels, and many large customers have no long-term contracts and can cut budgets at any time. Compute commitments are rigid and ten-year, while revenue is elastic and quarterly—this mismatch is the most concrete risk in the entire prospectus. Also do not forget the political variable: in March this year, the Department of Defense put Anthropic on the supply chain risk list (halted by a judge in August), and OpenAI has also publicly questioned its recognition of channel revenue on a gross basis and inflated run rate. These disputes will all be brought up one by one by institutions during the roadshow Q&A.

Conversely, Anthropic's willingness to open its doors at this time is itself a statement: the public market wants to see audited financial statements, customer retention and gross margin curves, so here they are. For the entire AI industry, the real significance of this IPO may lie here—for the first time, the valuation narrative of large models will have to withstand quarterly public-market scrutiny. If the answers are good, the pricing of OpenAI and xAI later will all have an anchor; if not, the debate over the "AI bubble" will immediately gain new ammunition.

Conclusion

From $380 billion to $965 billion and then to a run at $2 trillion, Anthropic's valuation completed a three-step jump in 18 months, and the story supporting it finally shifted from "how fast is growth" to "can profitability last."

The roadshow in the week of November 9 will be the first time a frontier AI lab has opened its books to the public market—an adjusted gross margin of 85% is beautiful, and a $518 billion compute commitment is also frightening; what Wall Street has to do is price between the two. The answer will be known before Thanksgiving. Channel: Finance and Economics This content is published with the author's authorization, and the views represent only the author and do not represent Huxiu's position. If you have objections or complaints about this manuscript, please contact tougao@huxiu.com. Everyone who is changing and wants to change the world is in the Huxiu APP[ ]]]

Source

Huxiu