Anthropic, the company behind the Claude AI models, has filed publicly to sell its shares on the stock market. In the registration statement it filed on October 26, 2026, Anthropic reports $16.3 billion in revenue for the first six months of 2026. In the same period a year earlier it took in roughly $1.2 billion.
Peter Diamandis, founder of XPRIZE and Singularity University, made that figure his "number of the week" when the Moonshots panel recorded the next day, October 27. He put the year-earlier revenue at $1.16 billion, which makes the jump a little over 14 times. "Dave, have you ever seen anything ramped this fast?" he asked.
Dave Blundin, founder and general partner of the venture firm Link Ventures, said he didn't think anyone had. "Coming from a cold start, it's got to shatter every record by far," he said, adding that anyone who claims to have seen this before "has never seen this movie before."
What the filing is
An S-1 is the document a company files with the US Securities and Exchange Commission before it sells shares to the public. It sets out the company's finances, who owns and controls it, and the risks a buyer would take on. Anthropic's applies to list on the Nasdaq exchange under the ticker ANT. The fields for the share price and the number of shares were still blank.
Diamandis filled in the rest from press reports. Citing the Wall Street Journal, he said the listing was likely in the week of November 16, with Anthropic raising $100 billion at a valuation between $1.5 trillion and $2 trillion. He said that would be the biggest IPO in history. None of those figures appears in the filing, so they are his account of the reporting, not the company's terms.
The strengths Diamandis read out
Diamandis went through the filing as a list of strengths and weaknesses. On the strong side, he said Anthropic earned $4.6 billion in all of 2025, so the first half of 2026 alone brought in more than three times that.
He also said that in the second quarter Anthropic made its first operating profit, of "nearly a billion dollars." Operating profit is what a company keeps from its revenue after paying its running costs. The filing's quarterly figures bear that out: it reports operating income of $974 million for the three months ended June 30, 2026, and every earlier quarter it lists, back to early 2025, was a loss. Because the first quarter of 2026 lost $1.67 billion, the six months as a whole still show an operating loss of $693 million, meaning costs exceeded revenue across the half. That half-year loss is much smaller than the $2.1 billion loss a year earlier.
The point Diamandis lingered on was capital efficiency, meaning how much business a company gets out of what it spends. Training and running AI models takes enormous amounts of computing power, which Anthropic mostly buys or rents. He said Anthropic now earns $2.08 in revenue for every dollar of compute, up from $0.40 eighteen months ago: "put in a dollar compute and double it in revenue." He also said 11,900 companies pay Anthropic more than $100,000 a year, which he took as a sign the company has spread its customer base.
The weaknesses
The first weakness Diamandis named was those compute bills. He said Anthropic has more than $500 billion in commitments to buy computing capacity, "much of it non-cancelable," including $120.5 billion with SpaceX. The filing describes agreements with SpaceXAI worth up to $120.5 billion through 2029. They are less fixed than "non-cancelable" suggests: once their initial locked-in periods end in January and March 2027, either side can cancel with 90 days' notice. Blundin made a similar point, saying the compute spend was non-cancelable "but with all kinds of knobs and bells in there."
The second was revenue concentration. Diamandis said two customers account for about 20% of first-half revenue, and nearly half of all revenue flows through the cloud platforms of Amazon, Google and Microsoft, which offer Anthropic's models to their own customers. He pointed out that those companies are also its competitors.
The third was control. Diamandis said the seven cofounders keep 50.1% of the votes through one special share, so public investors would not control the company. He allowed that "you can call it a weakness or a strength, depending on your point of view." In the filing, a company owned by the founders holds a single Class F share carrying 50.1% of general voting power. A separate Class T share governs how the board majority chosen by Anthropic's Long-Term Benefit Trust is elected.
Why Blundin would buy anyway
Blundin's answer to the list was that only the compute commitments give critics anything to poke at. "Compared to the Internet, nineteen ninety nine era," he said, "this company is just a cash machine." He didn't call the stock a bargain. A $2 trillion valuation "bakes in a lot of goodness," he said. "So that doesn't make it cheap. It just makes it the most extraordinary company in history by a wide margin."
His yardstick was Cerebras, the AI chipmaker, which he said went public perhaps a year to a year and a half earlier at a valuation of maybe $40 billion while raising a couple of billion dollars. On the reported terms, Anthropic would raise more money on its first day of trading than Cerebras was worth in total. "If the stock goes up, great. If the stock goes sideways, great. If the stock goes down by half, it's still the biggest thing ever," he said.
He turned that into advice for listeners: don't let an inflammatory headline slow down your life plans. "You need to move as fast as you possibly can," he said. He thought Anthropic was "correctly priced," and he called the super-voting structure "not that uncommon these days" but "a little bit weird."
Diamandis asked him directly whether he would buy into the IPO. Blundin said the question was hypothetical for him, since he has access to seed-stage deals, the earliest rounds of a startup's funding. But if he were an ordinary investor "just out there on the street," he said, "I would definitely buy in," adding that this was not investment advice. His reasoning was about exposure. AI is hard for most people to invest in directly, he said, and a listed Anthropic gives them a way in. Buying SpaceX would also do that, he noted, but SpaceX comes with many other businesses attached. "I wouldn't, you know, mortgage the house or anything like that," he said, "but I would definitely want exposure to the category."
Who got in early
The panel then looked at the people who funded Anthropic before it was famous. Blundin said that for Menlo Ventures, one of its investors, "this one deal dwarfs the entire history of the venture fund."
Diamandis said there were no venture capital firms in Anthropic's first two funding rounds, the Series A and Series B, and gave his own return figures. Anthropic's own announcements correct some of his details. The Series A, announced in May 2021, did raise $124 million, as he said. It was led by Skype cofounder Jaan Tallinn. Dustin Moskovitz and Eric Schmidt, whom Diamandis named alongside Tallinn, were participants. Diamandis said that round will return 345 times its money, about $43 billion for those early backers.
He put the Series B at $982 million, led by Sam Bankman-Fried and his crypto exchange FTX. Anthropic's April 2022 announcement gives a smaller figure: $580 million, led by Bankman-Fried, then FTX's chief executive. Diamandis said FTX's $500 million stake would now be worth $39 billion, about 79 times what it paid, but that the FTX estate sold it for $1.3 billion in 2024. "It's got to be the best trade ever made," he said.
Where the money goes next
The conversation turned to what the IPO money will do. The argument was that the real opportunity lies less in buying Anthropic shares than in backing early-stage investors like Blundin, because founders, employees and funds cashing out will put the money straight back into AI startups. Diamandis agreed, saying the Anthropic, OpenAI and SpaceX IPOs would "mint more billionaires" who would reinvest in the same ecosystem.
Blundin took it further. After these IPOs, he said, San Francisco becomes "overwhelmingly the financial capital of the entire world," ahead of New York and London. He noted that the city has only about 900,000 people, a tenth of New York, and said the concentration of wealth there "is going to be like nothing anyone's ever seen." Most of that money, in his view, wants to go back into "the machine that created the money, which is an AI machine." His advice to entrepreneurs was to rent an Airbnb, spend as much time in San Francisco as they can, and avoid becoming California residents.
"Not tax advice," said Alexander Wissner-Gross, a computer scientist and founder of Reified. Salim Ismail, founder of OpenExO, joked about another consequence of the boom: "They're going to have to import a bunch more hipster baristas in from Portland."