On October 15 I asked Claude a bar-room question. I wanted Anthropic’s and OpenAI’s quarterly revenue side by side and then, in the same chart, “the company that until today held the fastest revenue growth in history, to see who wins”. It suggested Nvidia and I stopped it, because I wanted “a company from the past that set revenue growth records, like Amazon or WeWork or something else, something old, not Nvidia”.
I know, it is like asking the innkeeper whether his wine is any good (an Italian saying): I asked Anthropic’s own product to measure Anthropic. That is why every number below was checked one by one against the original documents, from SEC filings to Reuters reports, and along the way the answer changed.
How much revenue does Anthropic make compared with OpenAI?
In the third quarter of 2026 Anthropic booked $16.8 billion of revenue, 45% more than the $11.6 billion of the second quarter. That is more than three times the $4.7 billion of the first. These are the figures the company shared with investors at an investor day, according to Reuters on October 14: numbers reported by people who were there, not yet published financial statements.
According to CNBC, OpenAI told investors it made $6.7 billion in the second quarter, 18% more than in the first, which puts the first at about $5.7 billion. So the overtaking happened in the spring, even though over the whole of 2025 OpenAI was still far ahead, with $13.07 billion (from leaked audited financial statements) against roughly $4.6 billion. We found no public trace of OpenAI’s third quarter.
Over the whole of 2025 OpenAI was ahead: $13.07 billion against roughly $4.6 billion. Sources: Reuters, 10/14/2026 (Anthropic, figures shared with investors, not published financial statements); CNBC, 8/19/2026 (OpenAI, figures shared with investors: $6.7 billion in Q2, up 18% on Q1, which gives about $5.7 billion); Quartz, 6/16/2026 (OpenAI 2025, leaked audited financial statements); Reuters and Fortune on the prospectus (Anthropic 2025). OpenAI's third quarter: no public trace found as of 10/15/2026.
Who held the record before Anthropic?
Comparing a company that bills billions with one that billed ten million makes no sense, because everyone grows fast from a small base. The rule Claude picked, and which I keep, is simple. For each company you start from the first year with revenue between $300 and $800 million and look at how many times revenue multiplies the following year.
Anthropic made about $386 million in 2024 and about $4.6 billion in 2025: 11.9x. The best in the chat’s sample was Groupon, from $312.9 million in 2010 to $1,610 million in 2011, or 5.1x. Then Google at 3.3x from 2002, Amazon at 2.7x from 1998, Facebook at 2.5x from 2009 and WeWork, the one I had named, at just 2x from 2016.
After two years the comparison stops being a race. With only nine months of 2026 Anthropic is already above 80 times its 2024, while Groupon and Google were around 7x after two years.
After one year
After two years
Anthropic after two years: only nine months of 2026, figures shared with investors and not yet in published financial statements (Reuters, 10/14/2026); 2024 roughly $386 million (Daily Caller on the prospectus; matches Fortune's +1,088%). Groupon at net revenue, as restated in 2011. Google at gross revenue, as reclassified in 2004. BioNTech in euros. Moderna skips the band ($60 million in 2019, $803 million in 2020) and starts from 2020. Sources: filings with the SEC (S-1, 10-K, 20-F). Nicola Giunchi's elaboration.
Against Nvidia, which Claude had suggested first, the result is the same. At the height of the AI boom, between spring and autumn 2023, Nvidia went from $7.19 billion to $18.12 billion a quarter in two quarters, 2.5x. Anthropic did 3.6x over the same number of quarters and added more dollars: $12.1 billion against $10.9 billion.
Start: Anthropic and OpenAI Q1 2026; Nvidia February-April 2023 quarter (Q1 of fiscal 2024). Sources: Reuters, 10/14/2026 (Anthropic, figures shared with investors); CNBC, 8/19/2026 (OpenAI; Q1 derived from the 18% growth into Q2); Nvidia, 10-Q and 10-K filed with the SEC (the fourth quarter is the 10-K minus the first three).
Was Groupon’s record real?
In its June 2011 prospectus Groupon reported 2010 revenue of $713.4 million, counting the full price of the vouchers it sold, including the share that went to merchants. The SEC pushed back, and in September Groupon restated its accounts “to correct for an error” in presentation: 2010 revenue became $312.9 million, the part it actually kept. More than half of the receipt was never Groupon’s, and the 5.1x in the chart is already calculated on the net figure.
Google had gone the opposite way in 2004. In its first prospectus, in April, it reported 2002 revenue of $348 million, net of what it paid to the sites in its advertising network. With an amendment at the end of July, confirmed in the final prospectus in August, it reported the same year at $440 million, gross, and called it a reclassification. The question underneath is always the same, who is really the seller and who is only an intermediary, but Groupon had to call its answer an error.
Anthropic has this question at home. According to Reuters, which read the prospectus, sales through the Amazon and Google marketplaces were 47% of revenue in 2025. Anthropic books them gross and records the platforms’ fee as a marketing cost. OpenAI has told its investors that this inflates Anthropic’s reported revenue by billions. The fees, as estimated by Reuters, come to about $351 million, 8% of 2025 revenue: take all of them out of 2025 and none out of 2024, and the multiple falls from 11.9x to about 11x. The record holds also because the share going to third parties is very different. According to Reuters, on marketplace sales the platforms keep some 16 cents for every dollar, against more than half of Groupon’s receipt.
So is it the fastest growth in history?
No, or at least not in year one. The chat left out two companies that break the chart. BioNTech went from €482 million in 2020 to almost €19 billion in 2021, 39x. Moderna skips that band, going from $60 million to $803 million between 2019 and 2020, and the following year reached $18.5 billion, 23x. They were the Covid vaccines, sold in the middle of a pandemic.
In 2023 BioNTech’s revenue was €3.8 billion, 80% below its peak. Over two years Anthropic beats them too, above 80x against BioNTech’s 36x. Yet the comparison teaches something the ranking does not say: growth like this can also come from a need that goes away. The right question, then, is whether Anthropic looks more like a pandemic or like a platform. By platform I mean a tool that companies build their own work on, and therefore keep paying for.
There is no shortage of warning signs on Anthropic. Third-quarter profit is “adjusted”, meaning calculated before some cost items. In 2025 two customers were worth 12% of revenue each, and Anthropic itself warns in the prospectus, according to Reuters, that many of its biggest customers have no long-term contracts. In August, in my piece on the AI bubble, I wrote that Anthropic was getting ready to list on a growth narrative of seven times in seven months, measured on run rate. Today that growth also shows in quarterly revenue, and the real exam will be the public prospectus, with statements signed by the auditors.
How much could Anthropic make in 2026?
If the fourth quarter grows 45% like the third, Anthropic closes 2026 at around $57 billion, within a range of $50 to $65 billion. This is our estimate on figures reported to investors, not an Anthropic figure.
Anthropic's figures: 2024 roughly $386 million (Daily Caller on the prospectus; matches Fortune's +1,088%) and 2025 roughly $4.6 billion (Reuters and Fortune on the prospectus, 9/29/2026), revenue for the year, not run rate; 2026, first three quarters $33.1 billion presented at an investor day and reported by two people who attended (Reuters, 10/14/2026), not financial statements. The fourth quarter is our calculation, not an Anthropic figure: it grows like Q3 (+44.8%), 16.8 × 1.448 = 24.3, so 2026 ≈ 33.1 + 24.3 = 57.4. Range from 49.9 (Q4 equal to Q3) to 64.9 (Q4 at the compound average growth, +89.1%): the upper end requires quarterly growth to climb back from +45% to +89%. Nicola Giunchi's elaboration.
Q1-Q3: figures presented at an investor day and reported by two people who attended (Reuters, 10/14/2026), not financial statements. The fourth quarter is our calculation, not an Anthropic figure: it grows like Q3 (+44.8%), 16.8 × 1.448 = 24.3. Range from 16.8 (equal to Q3) to 31.8 (compound average growth of Q1→Q2 and Q2→Q3, +89.1%): the upper end requires quarterly growth to climb back from +45% to +89%. Nicola Giunchi's elaboration.
What do I take away from this?
I use AI as a partner, not as a tool, and I described how in how I use AI. Those $16.8 billion a quarter are money companies pay to use it, so they will be the ones deciding between pandemic and platform, by renewing or not. In how we use Claude in Italy I went through the Anthropic Economic Index: among eight comparable countries, Italy ranks first for the weight of regulatory compliance and last for sales and strategy.
I started by asking who wins the race and I end with a question that is more useful for anyone running a business. If Anthropic’s AI is a platform, whoever uses it to sell gets there before everyone else, and in Italy, more than elsewhere, we use it for paperwork. Anthropic’s record will stay Anthropic’s; the record of an Italian company that grows by using AI to sell is still there to be written. What are you using it for?
Public sources and one estimate of our own: none of this is investment advice.
Sources: Reuters, third-quarter revenue (10/14/2026); Reuters, Anthropic’s prospectus (9/29/2026) and marketplaces and customers (The Star, 9/30/2026); Fortune (9/29/2026); Daily Caller (9/29/2026); CNBC, OpenAI (8/19/2026); Quartz on Yahoo Finance, OpenAI’s 2025 financials (6/16/2026); Groupon, S-1 (2011), S-1/A with the restatement (2011), 10-K 2011 and 10-K 2012; InvestorPlace on Groupon and the SEC (2012); Google, S-1 (April 2004), S-1/A (July 2004), final prospectus (August 2004) and 10-K 2005; Amazon, 10-K 2000; Facebook, S-1 (2012); WeWork, S-1 (2019); BioNTech, 20-F 2022 and 20-F 2023; Tesla, Moderna and Nvidia from the XBRL data of their SEC filings.