What did Anthropic's amended IPO filing disclose?
Anthropic filed an amended registration statement with the U.S. Securities and Exchange Commission on Nov. 2. The update adds third-quarter financials to the IPO prospectus it first published on Oct. 26. That earlier document included results only through the first half of 2026.
The new numbers are large in both directions. Revenue for the quarter ended Sept. 30 was $17.0 billion, up 1,127% from $1.382 billion a year earlier. Revenue for the nine months reached $33.24 billion, compared with $2.545 billion in the same stretch of 2025.
Profitability remains the harder half of the story. Anthropic reported a net loss of $23.29 billion for the third quarter. Its nine-month net loss was $104.04 billion, a figure dominated by a $94.3 billion accounting charge tied to financing agreements.
Those marks have built up over time. Anthropic reported an accumulated deficit of $156.2 billion, including $135.1 billion tied to changes in the fair value of convertible notes and derivative liabilities since 2023. Against that, operations generated $6.6 billion of cash in the first nine months of 2026. A year earlier, operations used $2.0 billion.
The filing also put a number on future spending. As of and after Sept. 30, Anthropic had purchase commitments and lease agreements with an aggregate committed spend of about $895.4 billion. The company said "substantially all" of that amount relates to compute and infrastructure.
Anthropic did not disclose how many shares it will sell or at what price. It has applied to list Class A common stock on the Nasdaq Global Select Market under the ticker ANT. The Associated Press reported that the company is preparing for a "blockbuster Wall Street debut later this month."
Why did Anthropic's revenue growth slow in the third quarter?
The quarterly ladder shows how quickly Anthropic has scaled. Revenue was $376 million in the first quarter of 2025, $787 million in the second, $1.382 billion in the third and $2.04 billion in the fourth. It then rose to $4.731 billion, $11.555 billion and $16.956 billion in the first three quarters of 2026.
Year-over-year growth is still extreme. The sequential trend is cooler. Third-quarter revenue rose about 47% from the second quarter. That compares with sequential growth of 144% in the second quarter and 132% in the first, according to CNBC's reading of the filing.
Anthropic expects the rate to keep falling. The filing says recent growth "may reflect early-stage adoption dynamics" and the nascency of the AI market. It also warns that revenue growth may decline significantly and faster than the company anticipates as the base gets larger.
The absolute numbers still moved fast. Revenue increased by about $5.4 billion from the second quarter, more than Anthropic's entire 2025 revenue of $4.585 billion. Most of the money comes from usage rather than subscriptions. Consumption-based revenue was $26.3 billion, or 79% of the nine-month total. Subscriptions contributed $6.9 billion, or 21%.
The customer mix explains the split. Customers spending at least $100,000 a year generated 74% of trailing-twelve-month revenue. The other 26% came from roughly 40 million customers who each spend less than that threshold. Growth therefore depends heavily on large accounts expanding, not just on new sign-ups.
The backdrop matters for the IPO. CNBC noted that investors are worried about rising interest rates, data center debt and the cash flow strain of hyperscaler AI spending. Those concerns could shape how the market values a company that is still losing money on a GAAP basis.
What does the $895.4 billion in purchase commitments cover?
The figure is not a cash balance or a loan. It is the total of contractual purchase commitments and lease agreements, including short-term commitments and leases that have not yet commenced. Anthropic says nearly all of it relates to compute and infrastructure.
The schedule is long. About $447.5 billion is expected to be incurred before Dec. 31, 2030. Roughly $69.8 billion is due in the 12 months after Sept. 30, 2026. Most of the remainder is expected to fall between 2031 and 2045.
A growing share comes from a strategy Anthropic calls Anthropic Direct. Instead of buying all capacity through fixed-term cloud contracts, the company is leasing chips directly and hosting them in dedicated facilities. It says this separates data center capacity commitments from accelerator investments. Data centers take longer to build but represent a smaller share of compute costs.
The approach has a near-term cost. In the third quarter, Anthropic delivered $5.0 billion of compute under short-term commitments. That capacity currently costs significantly more than comparable compute bought through multi-year deals. The company said this reduced revenue per dollar of compute compared with the prior quarter.
The scale is unusual even for AI. The commitments equal about 13 times annualized third-quarter revenue of $67.8 billion. They are roughly 6.8 times Anthropic's $131.9 billion of cash and short-term investments, and about 195 times its 2025 revenue. The company reported $146.6 billion of liquidity, including a $14.8 billion credit facility. It also cited up to $77.0 billion of available partner financing and $5.0 billion of equity commitments.
Some of the contracts carry minimum purchase commitments and take-or-pay provisions. Those terms obligate Anthropic to pay even if it uses less capacity than planned. The company says it secures capacity well before demand arrives because data centers take years to build. That timing mismatch is the central operational risk in the filing: too little capacity caps revenue, and too much raises costs.
Why is the net loss $23.3 billion when the operating loss is $567 million?
The two numbers measure different things. The operating loss reflects the cost of running the business: compute, staff, sales and administration. The net loss also includes interest income, taxes and changes in the value of financial instruments.
Most of the gap is non-cash. Anthropic's convertible notes, derivative liabilities and simple agreements for future equity are marked to fair value each period. Those marks cost $21.9 billion in the third quarter and $94.3 billion over nine months. A year earlier, the nine-month charge was $20.1 billion. The filing attributes the swings mainly to increases in Anthropic's enterprise valuation.
The operating line also moved in the wrong direction. Anthropic posted a $567 million operating loss in the third quarter, after $974 million of operating income in the second. The filing cites the $5.0 billion of short-term compute and $2.0 billion of non-cash charges linked to its donation matching program.
On the company's adjusted measure, the business was profitable. Adjusted operating income was $2.5 billion in the third quarter, up from $2.2 billion in the second. Adjusted operating margin fell from 19% to 15%, mostly because of the expensive short-term compute. The adjusted figure excludes stock-based compensation, non-cash charitable contributions and the Bartz settlement.
For the nine months, the operating loss was $1.26 billion, down from $5.22 billion a year earlier. Adjusted operating income was $4.1 billion, compared with an adjusted operating loss of $3.1 billion in the prior period. The headline net loss overstates cash burn. The operating result still shows a thin margin that depends on how quickly cheaper capacity comes online.
The cash flow statement offers another view. Operations generated $6.6 billion of cash in the first nine months of 2026, compared with $2.0 billion of cash used a year earlier. The company still reported negative operating and investing cash flows in 2023, 2024 and 2025. It expects the build-out of its own data centers to be a significant use of cash for the foreseeable future.
How large is Anthropic's business beyond quarterly revenue?
Anthropic counts customers by how much they spend on a trailing-twelve-month basis. As of Sept. 30, it had more than 11,900 customers spending at least $100,000 a year. More than 1,900 spent at least $1 million, and about 250 spent at least $10 million.
Large enterprises still devote a tiny share of their budgets to Claude. Every Fortune 10 company was a customer spending at least $100,000 on a trailing basis. So were more than 50% of the Fortune 500 and more than 30% of the Global 2000. Anthropic says those Fortune 500 customers spend less than 0.1% of their total expenses on Claude on average. Its top 10 Fortune 500 accounts spend under 1%, or about eight times that average.
The company also reports net revenue retention of approximately 900%. That is a company-defined metric. It divides the current revenue of a group of qualified customers by the revenue those customers generated in a base period. A figure that high signals rapid expansion inside accounts. It also reflects a small and fast-growing base, so it may not persist.
Revenue per employee has climbed with the business. Trailing-twelve-month revenue per employee was over $6.6 million on Sept. 30, compared with $1.4 million a year earlier. Revenue grew about 1,200% over that span while headcount rose about 170%.
Anthropic is also using its own models heavily. More than 90% of contributions to its codebase in September 2026 were written by Claude, the filing says. The average daily volume of code contributed per engineer in the third quarter was more than 45 times the average in 2024.
The product line has widened. Claude Code is the coding agent. Claude Cowork, a knowledge-work agent released as a research preview in January 2026, was merged with Claude.ai in September. The company also launched Claude Security and Claude Design. It sells four model classes: Mythos, which includes Fable, plus Opus, Sonnet and Haiku. Anthropic says Opus 5.5 cut customer costs per task by about 40% compared with Opus 5.
What does the filing say about AI speeding up AI research?
Anthropic describes a recursive process it calls "closing the loop." Claude helps design, train, test and optimize the next generation of Claude. The filing says this can compound capability gains. It also warns that a systemic flaw in one model could propagate into later versions built with its help.
The safety language is unusually direct for an IPO document. Anthropic says its internal estimates and expectations about future models suggest it may "imminently meet" its Responsible Scaling Policy definition of "dramatically accelerated progress attributable to AI." That threshold is based on forecasts, not current models. The company says it has not designated any current model as exceeding it.
Anthropic says it is still assessing the data and will publish further reports. It also says precautionary measures could slow releases relative to competitors. Past measures included delaying a model and limiting its initial availability, which the filing says happened with Mythos through Project Glasswing.
The company concedes that much of its evidence about acceleration comes from lagging indicators. It says it may not identify a jump in development speed until after it has happened. That uncertainty is why it says it has adopted precautionary measures in the past and may adopt more in the future.
The issue carries commercial risk, not just safety risk. If capabilities outrun the company's safeguards, Anthropic says it may face safety incidents, regulatory action or reputational harm. CNBC reported that the risk factors also cover political headwinds, complex infrastructure needs and tension between safety and commercial goals. The document includes more than 90 pages of such warnings. The co-founders wrote that they consider it "our personal responsibility to help the world understand these risks."
What happens next for the Anthropic IPO?
Anthropic is heading into its road show without a published price range. The amended filing leaves the share count and the estimated offering price blank, which is normal at this stage. AP reported that the debut is expected later this month, while CNBC said the company is preparing to begin its road show.
Part of the offering is aimed at Japan. Reuters reported that Anthropic plans to sell $3.5 billion, or about 552.9 billion yen, of shares to Japanese investors. The filing also includes selling stockholders, who will receive proceeds that Anthropic will not.
The corporate structure is unusual. Anthropic is a public benefit corporation, which requires it to balance shareholder returns with its stated mission and other stakeholders. A Long-Term Benefit Trust holds Class T stock and generally elects a majority of the board. A Founder LLC will hold Class F stock with 50.1% of total voting power. Class N stock is held mainly by strategic partners. The filing notes that Anthropic will be a "controlled company" under Nasdaq rules because of this structure.
The underwriting syndicate is long. Goldman Sachs, Morgan Stanley, J.P. Morgan and Citigroup lead a group that includes Barclays, Wells Fargo, BofA Securities, Deutsche Bank, RBC, UBS and more than a dozen others.
The market test will be about more than Anthropic. OpenAI confidentially filed its own prospectus in June and has signaled a 2027 debut, CNBC reported. Investors are already nervous about the financing behind AI infrastructure. Anthropic's first quarterly report as a public company, and the pace at which its committed compute arrives, will matter as much as the IPO price.
Watch three things after pricing. First, whether the company converts its $895.4 billion of commitments into revenue fast enough to justify them. Second, whether adjusted margins hold as short-term compute is replaced by owned capacity. Third, whether safety-driven release delays become a recurring drag on product launches.