At 11:50 on Thursday morning in New York, NBC News reported, shares of Anthropic, the maker of the Claude AI models, traded in public for the first time. They opened at $111.50, against the $83 a share that buyers in the offering had paid. By the close they had given back much of that jump, ending at $97.17, 17% higher than the offering price, CNBC and NBC News reported. It was a rough day to make a debut: "The S&P 500, the Nasdaq and the Dow Jones Industrial Average all closed more than 2% lower," NBC News reported.
Anthropic had priced the shares on Wednesday above the $70 to $78 range it had marketed, Fortune reported. It raised $81.3 billion, and on Friday its banks bought $12.2 billion more under their option for extra shares, bringing the total to $93.5 billion, according to CNBC, which called it "the biggest initial public offering on record." At Thursday's close, counting every share and option, the company was worth roughly $1.8 trillion, CNBC reported.
The listing made the headlines. The prospectus, the final version of the offering document Anthropic filed with the Securities and Exchange Commission on Thursday, tells the story behind them: the companies that sell Anthropic its chips and cloud computing are also among its biggest owners, and Anthropic has signed up to spend $900.9 billion, almost all of it on computing power.
The suppliers are the shareholders
Alongside the public offering, seven companies agreed to buy $26.0 billion of a separate class of stock at the same price: Google, Nvidia, AMD, Broadcom, Micron, Microsoft and TSMC, the Taiwanese chipmaker. Those Class N shares carry no vote, and the prospectus describes who holds them plainly: "The holders of our Class N common stock are primarily strategic business partners."
The ties run both ways. In July, Anthropic agreed to buy, lease or use AMD's Helios rack systems, built on its MI450 graphics chips, and AMD in turn committed to buy up to $5.0 billion of Anthropic stock, paid in steps as the chips are delivered and put to work, the filing says. The $2.5 billion AMD paid in this week's private sale counts toward that promise.
Two cloud giants are among the largest owners. After the offering, Amazon holds 14.6% of Anthropic's stock and Google 13.4%, according to the prospectus. Amazon also lent Anthropic money through notes that turn into stock at the listing, at a price of no more than $2.0928 a share for its 2023 notes and $5.0478 for its 2024 notes. The public paid $83.00.
Counting the public offering and the private sale, Anthropic expects net proceeds of about $104.8 billion, before the banks' extra shares.
Where the money goes
The number that stands out in the prospectus is the commitments. As of the end of September and since, Anthropic has signed purchase agreements and leases worth about $900.9 billion, "substantially all of which is related to compute and infrastructure." About $451.6 billion falls due before the end of 2030 and most of the rest between 2031 and 2045. About $70.7 billion is due within the 12 months after September 30.
Many of those contracts can't be cut back if demand falls short. They include "minimum purchase commitments, take-or-pay provisions, and other terms that obligate us to pay for resources regardless of actual utilization," the filing says.
For scale: Amazon spent $131.8 billion on property and equipment in 2025, Alphabet $91.4 billion, and Microsoft $115.9 billion in its fiscal year that ended in June, their own filings show: together $339.2 billion in a year, for three of the biggest cloud companies.
Source: SEC EDGAR XBRL data
On CNBC on Thursday, chief executive Dario Amodei said the total looks different next to the company's resources: "If you look at the amount of cash we have on the balance sheet and the amount of revenue that we're already making per year, suddenly those commitments don't seem so large," he said. He added that the company tries to "err on the low side" when predicting how much computing it needs, CNBC reported.
The growth, and the losses
The revenue behind that confidence has grown fast. Anthropic's quarterly revenue rose from $376 million in the first three months of 2025 to $17.0 billion in the three months that ended in September. For the first nine months of 2026, revenue grew by $30.7 billion, or 1,206%, from a year earlier. Customers paying it at least $100,000 a year went from more than 950 to more than 11,900.
The company also gives a "Run-Rate Revenue" of about $81 billion as of October 31, up $10 billion in a month. That is not a year of sales: it takes the last 28 days of usage fees and one month of subscriptions and assumes they continue "at the same rate for a full year." The prospectus itself warns that the measure "may not be indicative of our revenue for any future period."
Anthropic has lost money every year since it was founded, including $104.0 billion in the first nine months of 2026. Most of that is an accounting effect rather than spending: $94.3 billion came from changes in the fair value of convertible notes and similar instruments, which rose because the company itself became worth more. On its operating numbers, the gap is narrowing: in the third quarter, revenue rose 1,127% from a year earlier while spending on computing and infrastructure rose 382%.
The filing is just as clear that this pace won't last. "We expect our revenue growth rate to decline over time, and it may decline significantly and more rapidly than we anticipate as our revenue base increases and as the market matures," it says. It also records a $1.5 billion settlement of a copyright class action, Bartz v. Anthropic, approved by a court in July.
Who is in charge
Public shareholders get little say. The seven co-founders, including Dario Amodei and his sister Daniela, Anthropic's president and chair, are the only members of a company that holds a class of stock with no economic rights but 50.1% of the votes on most matters. A separate body, the Long-Term Benefit Trust, "generally has the power to elect and remove a majority of the authorized members of our board of directors." The founders' votes elect the rest. Anthropic is a public benefit corporation, legally required to weigh its mission alongside its shareholders' interests.
That mission is where the loudest criticism lands. More than 90 academics and AI safety experts signed a statement before the listing, CNBC reported, that says: "Anthropic warns that recursively self-improving AI could concentrate power and escape human control, yet races to build it first." The prospectus uses the phrase "closing the loop" for "a development cycle in which successive versions of an SI model assist in building the next, with the goal of compounding capability gains over time." SI is its term for super intelligence.
Amodei's answer, on CNBC, is that safety is the business case: "The company that builds the safest, that has trust of the customers and the enterprises of the world, I think that is the company that's going to be both best for society, and in the long run, I hope best for the performance of the company."
Our read
This is analysis, not advice. Anthropic's listing makes public a loop that was already there in private: chipmakers and cloud companies put money into Anthropic, and Anthropic has committed far more than that to computing and infrastructure, the things those companies sell. That is not proof of anything wrong. Amazon and Google have been investors for years, and the customers are real: a jump to more than 11,900 large accounts is hard to explain with supplier money alone.
But it does tie a large part of the AI industry to one company's growth. The $70.7 billion due over the next year is fixed; revenue, by Anthropic's own warning, will grow more slowly. If spending on computing keeps growing far more slowly than revenue, as it did last quarter, the commitments will look like what Amodei says they are. If revenue growth slows sharply first, the strain will show up not only at Anthropic but in the order books of the companies that just bought its stock. "There is still considerable uncertainty around how the competitive landscape and economics of generative AI will evolve," Morningstar analyst Malik Ahmed Khan told Fortune.
The first test comes with Anthropic's first quarterly report as a public company. On Thursday the market put a price on Anthropic. The prospectus shows how much it has already promised to spend.
