Anthropic's $16.8 billion quarter meets a cooler IPO valuation

On Wednesday, 14 October 2026, Anthropic opened its San Francisco headquarters to its most prominent current investors and to prospective buyers of its upcoming IPO. CEO Dario Amodei and CFO Krishna Rao walked them through numbers the company had never shown outside a small circle: third quarter revenue of $16.8 billion, up 45% from $11.6 billion in the second quarter, and an adjusted profit for the quarter.

The Financial Times was first with the figure. Reuters, Bloomberg, CNBC and Axios confirmed the core numbers through their own sources within hours. Then, on Thursday, 15 October, Reuters added the twist that makes this more than a victory lap: in an exclusive report, four people familiar with the matter said that some of the same investors are now weighing a valuation of around $1.5 trillion, well below the roughly $2 trillion that had been expected for the November share sale.

So the company that may become the largest IPO in history just showed that frontier AI can make money, and in the same week learned that the market wants a discount anyway. This article goes through what Anthropic reportedly disclosed, why growth slowing from "more than double" to "plus 45%" matters so much, what the compute numbers say about the business, and what to watch before the listing.

One caveat up front. Every figure below comes from people who attended or were briefed on a private investor meeting. Anthropic has not published audited financials, and it did not immediately respond to requests for comment from Reuters. Treat the numbers as reported, not as filed.

The numbers from the investor day

Here is what the reporting adds up to:

• Q3 2026 revenue: $16.8 billion. All outlets agree on this number.
• Q2 2026 revenue: $11.6 billion according to Reuters and CNBC, $11.5 billion according to the FT as summarized by Simon Willison.
• Q1 2026 revenue: $4.7 billion according to Reuters and CNBC, about $4.8 billion according to the FT.
• Growth: about 45% from Q2 to Q3. From Q1 to Q2, revenue had more than doubled (roughly +147% on the Reuters numbers).
• Profitability: profitable on an adjusted basis in Q3. The FT put the adjusted operating margin at 11%, excluding stock based compensation.
• Cash: Anthropic told investors it was cash flow positive in the quarter. Bloomberg's sources described positive free cash flow and a gross margin of 85%.
• Compute efficiency: $1.70 of revenue for every $1 spent on computing power in Q3, according to the FT.

Add the three quarters together and Anthropic has booked roughly $33 billion in revenue in the first nine months of 2026, on the Reuters figures. For comparison, Reuters reported last month that the company's revenue for all of 2025 was nearly $4.6 billion, a twelvefold jump from the year before, while it lost more than $8 billion on an operating basis (excluding writedowns tied mostly to earlier fundraising). A single quarter of 2026 is now more than three times the whole of last year.

There were softer details too. Reuters says Amodei reiterated the company's focus on safety but did not bring up the risk of human extinction, even though the IPO prospectus Reuters saw in September warns that advanced AI could pose "catastrophic or existential risks to humanity". We covered that warning in our piece on the risk section of Anthropic's IPO filing. Executives instead talked about how Anthropic uses its own AI inside the business, including to help design a custom chip. Attendees went home with caps, some of which simply read "thinking".

Why 45% growth disappointed some investors

For almost any other company, a 45% jump in one quarter would be extraordinary. The problem is the line it is drawn against. Between the first and second quarter, Anthropic's revenue grew by roughly two and a half times. Investors who priced the IPO around $2 trillion were, in part, pricing a continuation of that curve.

According to the Reuters exclusive, some attendees were "disappointed by the slowing sequential growth rate". Several of them discussed a valuation of about $1.5 trillion, three of the four sources said. That is a markdown of roughly $500 billion before a single share has traded.

A simple way to see what that means is to compare the valuation with the latest quarter. If you multiply Q3 revenue by four, you get $67.2 billion of annualized revenue. That is our own calculation, not a company figure, and it ignores further growth. On that basis:

• $2 trillion is about 30 times annualized Q3 revenue.
• $1.5 trillion is about 22 times annualized Q3 revenue.

Both are very rich multiples. The debate is not whether Anthropic is a real business; the numbers settle that. It is how long growth can stay this steep, and therefore how much of the future investors should pay for today.

What is slowing growth

The FT's reporting, as summarized by Yahoo Finance and BeInCrypto, points to several pressures at once.

Big customers building their own models. Two large Anthropic clients, Meta and Microsoft, now lean more on models they develop in house. When customers of that size shift workloads, it shows up quickly in a single quarter.

Less distillation revenue, on purpose. This is the most interesting detail. Anthropic told investors that distillation, where rival labs train their own models on a stronger model's outputs, accounted for 5% to 7% of revenue, and that the company had cut it back "materially" in the third quarter, according to four people quoted by the FT. Simon Willison did the napkin math: if 5% of Q2 revenue came from distillers, other labs spent over half a billion dollars in that quarter alone to learn from Claude. We wrote about the technical side of this in our article on Anthropic's distillation defenses. The investor day shows the business side: Anthropic is giving up real revenue to protect its models.

A price war. Anthropic and OpenAI have both cut prices as low cost Chinese labs add pressure. The FT mentions that Anthropic priced Claude Haiku 5.5 about 75% below its predecessor this month; see our coverage of Claude Haiku 5.5 and of the price war in business AI spending. Lower prices can grow usage, but they slow revenue per customer.

The compute math flipped in a year

The number that may matter most for the long run is not revenue at all. It is the ratio between what Anthropic earns and what it spends on computing power.

According to the FT, Anthropic made $1.70 in revenue for every $1 of compute in the third quarter. In 2025, the company reportedly spent $7.33 billion on compute against about $4.6 billion in revenue, which works out to roughly $0.63 per compute dollar. In other words, a year ago every dollar of chips and cloud time lost money before salaries, research and everything else. Now each compute dollar returns well over a dollar and a half.

Axios adds outside estimates that help explain why. Bloomberg Intelligence estimates that Claude Code alone reached an annualized revenue of $6 billion to $7 billion by late summer, and that Anthropic charges roughly 20% to 30% more than competing models. SemiAnalysis estimates that Anthropic's API business brings in about 90% of revenue while using 58% of the compute, implying a gross margin of about 92% for that segment before certain expenses. These are analyst estimates, not company disclosures, but they fit the picture of a business where enterprise and developer customers pay a premium and stay.

Axios also put the cash flow claim in context: Alphabet, long one of the most profitable companies in the world, turned free cash flow negative in its second quarter because of AI infrastructure spending. A much younger company in the same race says it is cash flow positive.

The spending that is already committed

The profitable quarter does not make the cost problem go away. In its IPO prospectus, Anthropic committed to at least $518 billion over roughly a decade for cloud services, computing and infrastructure, most of it hard to cancel. We broke that down in our article on Anthropic's $518 billion compute commitments. Reuters also reports that the prospectus showed a loss of $42 billion in 2025 on a full accounting basis. That is far larger than the operating loss of more than $8 billion, which excludes writedowns of liabilities tied mostly to earlier fundraising.

Amodei and Rao reportedly warned investors that margins and profitability could change as the company invests heavily in compute and in biomedical research. Axios lists the open risks plainly: large training costs and stock based compensation still weigh on underlying profit, and cheaper open weight models could pressure premium pricing over time.

That last point is not abstract. On the same Thursday, Meta released the open weights of Muse Spark Open, an 816 billion parameter model it calls the most capable US open weight model so far. Every capable open model is a reason for a price sensitive customer to ask whether the premium is worth it.

Run rates and the year end question

Where Anthropic ends 2026 depends on who you ask, and the spread is wide:

• $65 billion: the annualized run rate reported for the end of July.
• $67.2 billion: Q3 revenue times four (our calculation).
• About $74 billion: the year end projection from SemiAnalysis, as reported via the FT.
• About $110 billion: the year end annualized revenue that analysts who met executives expect, according to Reuters, and that investors expect according to The Wall Street Journal, as cited by CNBC.

Simon Willison noticed something else in the FT story: Anthropic appears to no longer lean on "annualized run rate revenue" in its investor communications and instead talks in quarters. That is a healthy change. Run rates are a snapshot multiplied by twelve, and they can swing with one large contract. Quarterly revenue is closer to what will eventually appear in audited statements.

Why the whole AI market is watching

Reuters describes the Anthropic IPO as a bellwether. The listing is expected to come before a possible OpenAI IPO next year, which makes Anthropic the first real test of how public investors value a frontier lab that is spending hundreds of billions on compute.

The timing is awkward. On the day of the investor meeting, the S&P 500 fell by more than 1% for the first time in 53 sessions, according to Axios. On Thursday, Reuters reported that shares and bonds of Nvidia backed CoreWeave tumbled amid concern about an AI buildout increasingly financed with debt. Reuters also notes recent IPO setbacks: data center operator SB Energy delayed its listing last month and Holtec Nuclear pulled its plans.

A cooler reception for Anthropic would likely spill over to chipmakers, cloud providers and the neoclouds that rent out GPUs, because so many of their long term contracts depend on labs like Anthropic and OpenAI keeping their spending promises. A strong debut would do the opposite.

For context on the rival, OpenAI told investors earlier this month that its annualized revenue was approaching $50 billion at the end of September, with a different, net based accounting method for partner sales. We explained why the two companies' numbers are hard to compare in our article on OpenAI's revenue figure.

Timeline to the listing

• September 2026: Reuters sees the confidential IPO prospectus, including the $518 billion compute commitments and the existential risk warning.
• 13 October: CNBC reports that Anthropic will disclose third quarter results to potential investors for the first time.
• 14 October: Investor day in San Francisco with Amodei and Rao. The FT reports $16.8 billion; Reuters, Bloomberg, CNBC and Axios follow.
• 15 October: Reuters reports that some investors discuss about $1.5 trillion instead of about $2 trillion.
• November: the share sale that investors have been preparing for. Reuters has reported that the debut could slip until after the US midterm elections; Bloomberg's sources expect it no later than the end of the year.

What to watch next

1. The public filing. Once Anthropic files publicly, the quarterly numbers, the gross versus net treatment of cloud partner sales, the definition of "adjusted" profit and the stock based compensation bill will all be visible and audited. 2. The Q4 trajectory. If growth holds near 45% a quarter, the $110 billion year end figure comes into view. If it slows toward the SemiAnalysis path, the $1.5 trillion camp gets stronger. 3. Pricing. Watch whether the 20% to 30% premium survives the Haiku 5.5 price cut, cheaper Chinese models and strong open weights from Meta. 4. Compute ratio. The jump from $0.63 to $1.70 per compute dollar is the clearest sign that frontier AI can be a sustainable business. Whether it holds while the $518 billion in commitments ramps up is the real long term test.

Dany's take

I think both headlines are true at the same time, and that is the interesting part. Anthropic just showed the strongest evidence yet that a frontier lab can earn more than its chips cost, which a year ago was far from obvious. At the same time, a $1.5 trillion valuation is still an enormous bet on growth that only a few months ago looked unstoppable. Investors asking for a lower price after a profitable quarter is not a sign of weakness. It is what a market looks like when it starts to treat AI labs like businesses instead of stories. I would rather see that discipline now than after the listing.

The bottom line

Anthropic reportedly told investors that third quarter revenue rose 45% to $16.8 billion, with an adjusted profit, positive cash flow and $1.70 of revenue for every dollar of compute. Growth slowed from more than doubling in the previous quarter, distillation revenue was cut on purpose, and big customers are building their own models. Some investors now discuss a valuation of about $1.5 trillion instead of $2 trillion for the November IPO. Audited numbers will only come with the public filing.

Sources: Reuters: Investors temper Anthropic IPO valuation expectations over revenue growth slowdown (15 Oct 2026, via CNA), Reuters: At Anthropic pre-IPO meet (14 Oct 2026, via Investing.com), Financial Times: Anthropic revenue jumps to $16.8bn in third quarter (paywall), Bloomberg: Anthropic Revenue Hits $16.8 Billion in Quarter Before IPO, Bloomberg Law: Anthropic Results May Aid IPO But Not AI Trade, CNBC: Anthropic's Q3 revenue is said to hit $16.8 billion ahead of IPO, Axios: Anthropic brings Wall Street an AI bull case, Simon Willison on the FT details, Yahoo Finance / BeInCrypto on the compute ratio and growth pressures. All financial figures are as reported by these outlets from people familiar with the matter and were not independently verified.