Anthropic Shows Investors a $16.8 Billion Quarter; Growth Slows

SAN FRANCISCO, Oct. 17 — AnthropicAnthropic told prospective investors on Wednesday that its revenue in the third quarter was $16.8 billion, up about 45 percent from the second, and that it was profitable on an adjusted basis, two people who attended a closed investor day at its headquarters here told Reuters.

Within a day, attention had moved to the rate. Some current and prospective investors were disappointed by the company’s most recent revenue growth and had begun to discuss a valuation of about $1.5 trillion for its planned stock offering, rather than the roughly $2 trillion they had expected, Reuters reported on Thursday, citing four people familiar with the matter.

The sale could be among the largest ever, and would give public investors another direct way into a leading frontier AI lab. According to The Next Web’s account of Bloomberg’s reporting, it could match in size the record set in June by SpaceX, which sold $75 billion of stock. Reuters described Anthropic as “an early test of investor appetite for loss-making AI developers.”

The Quarters

The people at the meeting gave the year’s path as $4.7 billion of revenue in the first quarter, $11.6 billion in the second and $16.8 billion in the third, Reuters reported. The Financial Times, which first reported the quarter, put the first two slightly differently, at about $4.8 billion and $11.5 billion.Run RateQuarterly revenue is money earned in the period. A run rate annualizes a recent stretch. Anthropic said in May that its run rate had crossed $47 billion; analysts who met executives earlier in the week estimated it could reach $110 billion by year’s end, Reuters reported. On Reuters’ figures, revenue grew nearly two and a half times from the first quarter to the second, then by about 45 percent; in dollars, the third quarter added $5.2 billion, less than the $6.9 billion the second had added.

The first quarter alone exceeded Anthropic’s revenue for all of 2025, which Reuters, citing a prospectus it saw in September, put at nearly $4.6 billion, up twelvefold. The company lost more than $8 billion on an operating basis that year, excluding writedowns of liabilities tied mostly to earlier fundraising. Reuters reported a 2025 net loss of $42 billion, including a roughly $34 billion accounting charge from the rising estimated value of financing that could convert into shares, rather than money spent running the business. It also reported $7.33 billion of computing and infrastructure expense in 2025, and that two unnamed customers supplied nearly a quarter of that year’s revenue.

The largest number is what Anthropic has promised to spend: at least $518 billion over a decade for cloud services, computing and infrastructure, Reuters reported from the prospectus.ScaleAt the third quarter’s pace, about $67 billion a year, those commitments equal more than seven and a half years of revenue. Broadcom, named by Anthropic in May with Google as the source of five gigawatts of next-generation TPU capacity, disclosed in its quarterly report in September that a customer may issue it convertible notes of up to $42 billion, to be used solely for that customer’s lease obligations; Reuters, citing the prospectus, reported on Oct. 1 that the customer is Anthropic.AVGONasdaq326.08▼18.92−5.5%Broadcom · Friday’s session The facility is conditional; Reuters reported that some defaults could accelerate lease payments while limiting access to it.

The quarterly figures come from people who attended; the company has not publicly released them in financial statements. Reuters did not define the adjusted measure or give a dollar profit figure. Bloomberg, citing people familiar with the matter, reported an adjusted operating margin of over 11 percent; the Financial Times reported in September that the adjusted measure strips out costs including stock-based compensation. Bloomberg reported on Oct. 9 that Anthropic counts sales through cloud partners such as Amazon gross as revenue. Nor has it published a full reconciliation of the adjusted profit measure. Anthropic did not immediately respond to a Reuters request for comment and declined to comment to Bloomberg.

The Price

The last valuation Anthropic announced was set on May 28, when it raised $65 billion in a Series H round led by Altimeter Capital, Dragoneer, Greenoaks and Sequoia Capital, at a post-money valuation of $965 billion.The May RoundThe $65 billion included $15 billion previously committed by cloud companies, $5 billion of it from Amazon. Amazon and Google, both investors, also supply much of the computing Anthropic uses to train and run Claude. At $1.5 trillion, the company would be valued about 55 percent above that round; at $2 trillion, a little more than twice it. The $1.5 trillion figure is some investors’ discussion, not an announced company price. Reuters had reported that Anthropic was targeting a valuation above $2 trillion, but the company has not publicly announced an IPO price; when it disclosed its confidential filing on June 1, it said that “the number of shares to be offered and the price have not yet been set.”

On Wednesday, Mark D. Levine, the New York City comptroller, who is a trustee of city pension funds holding more than $325 billion, wrote to Dario Amodei, Anthropic’s chief executive, urging a pause. “Instead of proceeding headlong with its IPO, the company should focus its energies on addressing these unprecedented risks,” he wrote, citing reports that the draft prospectus warns advanced AI could pose “catastrophic or existential risks to humanity.” He asked for specific disclosure of planned audits and controls and a timeline, and warned that a governance structure described in press reports could concentrate control with founders and trustees. Anthropic has since opened a line of communication with Mr. Levine, Bloomberg reported on Friday; he told Bloomberg Television he could not characterize the discussions.

Instead of proceeding headlong with its IPO, the company should focus its energies on addressing these unprecedented risks.

The Calendar

Anthropic said on June 1 that it had confidentially submitted a draft registration statement on Form S-1 to the Securities and Exchange Commission. No public version had appeared on the SEC’s EDGAR system by Saturday evening, and Reuters has reported that the debut is likely to wait until after the Nov. 3 midterm elections. Bloomberg has reported, according to The Next Web, that formal marketing could begin as early as the week of Nov. 9, with trading before Thanksgiving, Nov. 26, and that the plans remain fluid.

SEC staff review a draft in private only if the company agrees to file it publicly at least 15 days before any road show. If the road show starts on Monday, Nov. 9, that points to a public filing by Oct. 25, a Sunday, or in practice Friday, Oct. 23; a later start would move the date. Renaissance Capital, the IPO research firm, wrote on Friday that the latest reports pointed to a listing “before the end of the year rather than firmly in November.”

On Thursday, Anthropic said it had given six employees of METR, a nonprofit that evaluates frontier AI models, “employee-like access,” with company hardware, office space and internal tools. Anthropic says METR will publish its assessments without a company veto, but Anthropic can require redactions and ultimately decides what falls within the permitted categories; METR reserves the right to make certain statements about the access and information it received.METRA nonprofit research group based in Berkeley, Calif. When the Federal Trade Commission confirmed its inquiry into AI companies on Sept. 30, it planned to request information from METR as well, CBS News reported. The company tied the step to its pledge under the White House Accord on Super Intelligence and wrote: “We’re embedding evaluators voluntarily, and we believe every frontier AI company should do the same.” On Wednesday, the Federal Trade Commission sent Anthropic and OpenAI civil investigative demands, CBS News reported, an inquiry covered elsewhere in this edition.

A public S-1 would put the company’s required audited annual financial statements in front of every investor, not only those invited to San Francisco. A filing by about Oct. 23 would fit a Nov. 9 road-show start; a later road show or listing would move that milestone.

Daily Cortex is written with AI models, including Anthropic’s Claude, and fact-checked with OpenAI’s GPT models.