Anthropic’s performance is “good but not good enough”: under “adjusted profitability,” revenue growth sharply brakes!

AI unicorn Anthropic has released a Q3 earnings report that looks impressive on the surface, but actually disappointed the market.

According to the Financial Times on Thursday, Anthropic disclosed during a closed-door pre-IPO Investor Day in San Francisco that its Q3 revenue reached $16.8 billion, up nearly 50% from $11.5 billion in Q2, and achieved "adjusted profitability" and positive cash flow.

However, the market reaction was far from upbeat. Market research firm P Equity Research stated bluntly: "Anthropic’s numbers are good, but not good enough." Investment analyst Geiger Capital was even more direct: "Not good enough. That's it."

The concern behind the numbers is the sharp slowdown in growth. According to ZeroHedge’s calculations, Anthropic's annualized revenue run rate had already reached $65 billion by the end of July this year, while Q3 data implies the monthly growth rate has plummeted from a previous compound growth of around 33% to as low as 3.4% to 6.6%. If this trend continues, the company’s annualized revenue by year-end would be about $89 billion, roughly $10 billion short of the previously widely expected year-end "over $100 billion" target.

This means that for Anthropic to achieve its own targets, Q4 monthly growth must accelerate again to 10%–13%—and this must happen even as token prices hit all-time lows and competition from open-source models continues to intensify.

Cliff-like Slowdown: From 33% Monthly Growth to Single Digits

The key to understanding this performance is reconstructing the true trajectory of monthly growth.

Anthropic customarily calculates annual run rate by multiplying the latest monthly revenue by twelve. By this measure, a $65 billion run rate at the end of July implies about $5.4 billion revenue for July. Back-calculating from Q3’s total revenue of $16.8 billion, combined revenue for August and September was about $11.4 billion, resulting in a month-over-month growth rate of only about 3.4%.

Even using a more generous assumption—treating the $65 billion as the instantaneous annualized run rate on July 31 rather than July’s average—July revenue would be about $5.2 billion, and the implied monthly growth rate for August and September would still be only 6.6%. By this calculation, Anthropic’s annual run rate at the end of Q3 would be between $69 billion and $74 billion.

This stands in sharp contrast to the growth trend in the first half of this year. From an annualized run rate of about $10.8 billion at the end of 2025 to $65 billion by the end of July, Anthropic achieved a compound monthly growth rate of about 33%. But by August and September, that number had fallen to single digits.

"Adjusted Profitability": Profits Full of Asterisks

Anthropic’s profitability claim deserved the most attention, and warrants the most caveats.

The Financial Times report did not reveal which items were actually “adjusted” out; the complete GAAP financials are expected to be disclosed with the public prospectus (S-1) around October 25. According to Reuters’ previous report on a confidential version of the S-1, Anthropic recorded a net loss of about $42 billion in 2025, including about $34 billion in financing-related accounting charges and over $8 billion in operating losses, with future commitments on compute and infrastructure over the next decade as high as $518 billion, of which about 80% are irrevocable commitments.

In addition, there is controversy around the definition of revenue itself. As explained by Goldman Sachs’ TMT team, Anthropic uses a near "gross revenue" definition—counting all customer spend through cloud partners, then recording partners' revenue share as costs. This makes its top-line numbers look better compared to OpenAI’s net revenue approach, and means "adjusted" profit margins are highly dependent on the specific accounting treatment.

With the largest compute bills not yet due (Microsoft commitments don’t start until November) and the IPO window about to open, single-quarter adjusted profitability is still a long way from a credible track record of profitability.

Token Prices Hit New Lows; Volume-Price Divergence Is the Core Contradiction

The structural root cause of the growth slowdown points to the sustained decline in token prices.

According to the Ornn Anthropic Token Price Index, Anthropic’s token prices have reached all-time lows. Meanwhile, the latest token usage data from Ramp AI shows that although AI usage is still growing rapidly (consistent with the Jevons Paradox), actual AI expenditures have at best flattened or even shrunk—the root cause being that the growing market share of open-source models has depressed overall token prices.

This volume-price divergence is now Anthropic’s core business model dilemma: The speed at which token prices are falling is faster than the company's ability to convert higher usage into revenue growth. If the August–September slowdown occurred while prices were still relatively firm, Q4’s required acceleration would have to rely on dramatically higher usage even as prices are cut in half—a fundamentally different business logic.

Meta, Google, and Chinese open-source heavyweight models are flooding the leading market with offerings at deep discounts, further compressing Anthropic’s pricing power.

For AI Trading: Good, But Not Good Enough

Whether these results support the overall AI investment narrative depends on the measurement standard chosen.

According to Goldman Sachs’ trading desk, citing Brad Gerstner’s framework, top AI labs need to raise their combined annual run rate from around $100 billion in July to between $180 billion and $200 billion by year-end to sustain the integrity of the AI trade. Based on Q3 trends, Anthropic will be around $89 billion at year-end, and combined with OpenAI’s target of about $70 billion, the two add up to only $159 billion—still below Gerstner’s threshold. And this is a mixture of gross revenue and net revenue definitions.

In terms of valuation, using a $2 trillion IPO valuation, Anthropic corresponds to about 27 times the implied run rate at the end of September, and about 22 times the December trend number—lower than Breakingviews’ estimate of about 47 times historical revenue just two weeks earlier. This is the bullish case investors heard on Investor Day.

However, the market voted with its feet the day before the earnings release: the S&P 500 Index fell 1.11%, the Nasdaq 100 fell 2.02%, sentiment in AI-related sectors declined in a “stair-step” fashion, emerging cloud-computing names sank more than 12% in a single session, CoreWeave shares plunged over 15%, and its CDS market implied five-year default probability has climbed to 63%.

$16.8 billion in quarterly revenue, adjusted profitability, and positive cash flow are enough to keep the November IPO plan on track. But the reality of a growth slowdown from 33% monthly growth to single digits, and the need for renewed Q4 acceleration to achieve the $100 billion year-end goal, are the real questions the market will need to answer once the S-1 is made public.