Anthropic publicly filed for an IPO with the U.S. Securities and Exchange Commission on Oct. 26 and plans to list on Nasdaq under the ticker ANT. The company is targeting a valuation of $1.5 trillion to $2 trillion, and media reports said the IPO window is set for the week of Nov. 16.
The filing landed as U.S. semiconductor stocks moved higher. The Nasdaq rose more than 1% intraday, while Broadcom, AMD and Nvidia each gained nearly 3%. Before Anthropic shares begin trading, investors are already reacting to the suppliers behind its expansion. The company has signed compute contracts worth more than $800 billion in total, with the longest commitments running through 2045.
The S-1 runs more than 300 pages and roughly 240,000 English words. It is the first full public look at Anthropic’s finances and governance structure. The Financial Times described the offering as a kind of referendum on whether public-market investors accept the narrative Dario Amodei used to win over private backers.
Revenue expanded at an exceptional pace
Anthropic reported $41 million in revenue in 2023, $386 million in 2024 and $4.585 billion in 2025. In the first half of 2026 alone, revenue reached $16.286 billion.
Quarterly figures show how quickly the business scaled. Revenue was $376 million in the first quarter of 2025, followed by $787 million, $1.382 billion, $2.040 billion and $4.731 billion in the next four quarters. It then reached $11.555 billion in the second quarter of 2026, about 30 times the level of five quarters earlier. That works out to roughly $127 million in revenue per day during the quarter.
For the first nine months of this year, media reports said Anthropic generated more than $33 billion in revenue, up from $4.6 billion a year earlier. Revenue per employee over the past year was reported at $6.6 million. Sales and marketing expense rose with the business, from $568 million in full-year 2025 to $3.168 billion in the first half of this year.
Growth, though, has started to slow. Revenue more than doubled from the first quarter to the second quarter, but the increase from the second quarter to the third quarter was closer to 50%. Anthropic warned in the filing that its rapid growth to date "may reflect early adoption" and could stall in the future.
A first-half net loss of $80.748 billion was driven mostly by non-cash items
The headline loss in the filing is a first-half net loss of $80.748 billion. Of that, $72.388 billion came from fair-value changes tied to convertible notes, derivative liabilities and SAFE instruments.
The mechanism is straightforward. Part of Amazon’s investment came in the form of convertible notes, effectively IOUs that can later convert into equity at agreed terms. Anthropic’s valuation was $380 billion in February and rose to $965 billion in May. As that valuation increased, the notes gained value, and accounting rules required the company to record the change as a loss. The filing says these changes were mainly driven by the higher valuation and did not involve cash outflows.
On Anthropic’s adjusted, non-GAAP basis, first-half adjusted operating profit was $1.618 billion, mainly excluding stock-based compensation and non-cash charitable donation expense. Adjusted net income was $2.671 billion.
Compute costs fell sharply relative to revenue
The Financial Times reported that in the first quarter of 2025, Anthropic spent more than $2.50 on compute and infrastructure for every $1 of revenue. By the second quarter of 2026, that figure had dropped to less than $0.50.
Anthropic’s annual figures point in the same direction. Compute spending was more than 10 times revenue in 2023, 653% of revenue in 2024, 160% in 2025 and 55% in the first half of 2026.
Margins improved with that shift. For the three months ended June 30, Anthropic posted an operating margin of 8% and operating profit of about $1 billion, compared with a $1.2 billion loss in the same period a year earlier. In the second quarter alone, operating profit was $974 million. Adding back stock-based compensation and other non-cash items, adjusted operating profit was $2.235 billion, with a 19% margin. Net cash provided by operating activities in the first half was $3.9 billion.
The filing gives four reasons:
- Training efficiency improved, with Claude increasingly involved in developing the next generation of models, a process the company calls "closing the loop".
- Inference became more efficient, reducing the compute needed for the same level of capability.
- Anthropic trained and ran models across four chip families at once — Amazon Trainium, Google TPU, Nvidia GPU and AMD GPU — while building its own scheduling system.
- New models became cheaper. After Opus 4.5 launched in November 2025, pricing and per-task costs fell, and first-quarter 2026 sequential revenue growth accelerated to 132% from 48% in the prior quarter. After Opus 5.5 and Fable 5.1 launched in September this year, per-task costs fell another 25% to 45%.
The company also warned that compute spending could create "significant period-to-period fluctuations" in results.
Customer concentration remains meaningful, but the mix is broadening
In the first nine months of the year, Anthropic’s top five customers accounted for 25% of revenue. Only 250 customers spent more than $10 million annually.
The concentration trend has been moving lower. The top five customers represented about 36% of revenue in the first quarter and about 21% in the third quarter. In the third quarter, 79% of revenue came from customers outside the top five, and no single customer accounted for more than 10%. At the same time, revenue from the top five customers in absolute terms rose from about $1.7 billion to about $3.5 billion. The filing also says about 48% of revenue came through cloud providers’ marketplace channels.
Customer counts expanded quickly. Customers spending more than $100,000 annually rose from about 20 at the end of 2023 to about 950 in September 2025 and more than 11,900 in September 2026. Anthropic also had about 40 million customers spending less than $100,000. More than half of Fortune 500 companies are already customers spending at least $100,000 a year, but their average spend on Anthropic is less than 0.1% of total spending, and even the top 10 users are below 1%.
Net revenue retention was about 900%, meaning a customer that spent $1 a year earlier spent an average of $9 in the comparable period this year. That figure excludes subscribers.
Usage-based revenue was $3.8 billion in 2025, or 83% of total revenue, and $13.1 billion in the first half of 2026, or 80%. Subscription revenue in the first half was $3.2 billion, up from $180 million a year earlier. More than 80% of third-quarter revenue came from model versions released within the prior six months.
Between 5% and 7% of revenue came from customers later found to be in violation
Anthropic said that for the nine months ended Sept. 30, about 5% to 7% of revenue came from customers later determined to have violated usage policies or commercial terms. The company said there is a lag between detection, investigation and account termination, and revenue from services provided before suspension is still recognized.
Anthropic said it stepped up identification and enforcement in August and September. It expects the share of such revenue to decline, but said that could also weigh on future revenue and growth. Based on roughly $33 billion in revenue for the first nine months, 5% to 7% would equal about $1.6 billion to $2.3 billion.
Compute commitments exceed $800 billion and stretch to 2045
Anthropic calls compute a "critical input" and says it has signed commitments worth more than $800 billion in total, with the longest extending to 2045. The founders acknowledged that these commitments may "at first glance seem irresponsibly large" and described the approach as "bold but rational."
The main contracts include:
- Broadcom: Broadcom and Google jointly develop TPUs and lease them to Anthropic under a five-year agreement worth $161 billion. Starting in 2028, Anthropic expects to spend up to another $150 billion on TPUs.
- AMD: expected spending of more than $20 billion, with options that could roughly double the scale. AMD will invest up to $5 billion in Anthropic as chips are delivered.
- Google, Amazon and Microsoft: at least $269.4 billion in cloud compute over the next 10 years.
- Nscale and Lambda: contracts worth up to $86 billion.
The three cloud providers have different timelines. Google will provide at least $114.5 billion from April 2026 to July 2033. Amazon will provide at least $110 billion through April 2036. Microsoft will provide at least $44.9 billion through May 2033. All three contracts include payment terms "regardless of actual usage," and the Google and Amazon agreements require Anthropic to make up the difference if usage falls short.
Using the notes to the financial statements, the filing shows $265.6 billion of fixed or minimum purchase commitments as of June 30, $321.4 billion of signed but not yet commenced data-center and compute leases, and $245 billion of new and amended long-term contracts signed after June 30. That totals more than $830 billion. As of June 30, Anthropic held about $131.6 billion in cash and short-term investments and raised about $110.4 billion through its Series G and Series H rounds in the first half.
The timing mismatch matters. About 80% of revenue is usage-based, while compute costs are often locked in more than a year in advance.
The SpaceX contract could reach $120.5 billion
Anthropic’s contract with SpaceX has a maximum value of $120.5 billion and runs through 2029. Once all compute capacity comes online in December, monthly payments are expected to be about $3.1 billion. Starting early next year, either side can terminate with 90 days’ notice.
The relationship shifted several times this year. In January, Anthropic cut off xAI employees’ access to Claude through Cursor. In February, after Anthropic announced financing, Elon Musk publicly criticized its AI. In May, Anthropic said it would lease xAI’s Colossus 1 data center in Memphis, and Musk later praised Anthropic’s people and said they care about doing the right thing. The filing names the counterparty as SpaceX.AI LLC, formerly X.AI LLC, now a wholly owned subsidiary of SpaceX.
According to SpaceX’s filing, its entire AI business generated $818 million in revenue in the first quarter of this year. Anthropic’s monthly compute bill would be nearly four times that figure.
Amazon and Google stakes come into view
The Financial Times estimated that if Anthropic lists at a $2 trillion valuation, Amazon would own about one-fifth of the company after full conversion of its notes, worth about $400 billion. Amazon’s ownership cap is 12%, with any excess retained in note form and converted over time as room opens up, though Amazon can waive that cap.
Google owns about 7% and may invest up to another $20 billion under an agreement reached earlier this year. Of that amount, $10 billion can be subscribed at the offering price. If both sides agree, Google could also act as an IPO cornerstone investor.
Among financial investors, the filing shows Spark Capital invested less than $1 billion and holds a stake valued at $14.5 billion. According to one informed person, Coatue invested $4.75 billion and holds a stake of similar size.
The lock-up structure is also unusual. As with SpaceX, Anthropic will have five lock-up release dates within 76 days after listing. Non-founder directors, executives and some current and former employees will be able to sell part of their shares on the first day of trading. Early investors in a typical IPO often wait 180 days before selling.
For employees, the filing includes a simulation based on a June 30 listing date. Total incentive expense would be about $29.7 billion. Spread across roughly 5,300 employees, that comes to at least $5.6 million per person on average. In an internal secondary sale in April, insiders including the Amodei siblings and the CFO sold $4.66 billion worth of stock.
Seven co-founders hold 50.1% of the voting power
Anthropic says in the filing that its "novel" governance structure differs materially from that of most public companies and leaves ordinary shareholders with almost no control over company decisions.
Media accounts describe a structure topped by a Long-Term Benefit Trust, or LTBT, whose members appoint themselves and generally have the power to elect and remove four of the seven directors. Dario Amodei and Daniela Amodei cannot be removed by the trust. The other three directors are nominally elected by shareholders, but in practice are determined by the two siblings and five other co-founders because they control 50.1% of the voting power through a special class of stock. The trust’s control over the fourth seat requires approval from a nominating committee that includes Daniela Amodei, and if there is a deadlock, an employee representative casts the tie-breaking vote.
The key instrument is a single Class F share that will be issued before the IPO to a Founder LLC made up of the seven co-founders. The share has no economic rights, but carries voting power equal to 50.1% of all voting power, including the election of all directors other than T directors.
By contrast, the Class N shares held by Amazon, Google and other large investors carry no voting rights. The filing also says Anthropic qualifies as a "controlled company" under Nasdaq rules and can claim exemptions from some governance requirements.
Philanthropy, risk disclosures and the founders’ warning to investors
Before the IPO, Anthropic had already allocated 846,840,960 Class N shares to its wholly owned Anthropic Philanthropic Fund, equal to 5% of fully diluted share capital. The board also approved a framework under which the company may contribute up to another 2% of shares to the fund each time the stock rises 50% from the prior milestone after the IPO. Each of the seven co-founders has pledged to direct 80% of personal equity to philanthropy.
The risk-factor section runs more than 90 pages, about three times the length of comparable disclosure in SpaceX’s filing. Anthropic warns that its technology could pose "catastrophic or existential risks" to humanity and could "provide substantial assistance to malicious actors seeking to develop chemical, biological, or other dangerous weapons."
The filing describes scenarios in which future models could show "unexpected ... self-preservation behavior," including attempts to resist shutdown, conceal or manipulate information, and produce coercive or deceptive outputs resembling "blackmail." On insurance, Anthropic says no single insurer, and possibly not the insurance industry as a whole, may be able to provide adequate coverage.
The company also says it is increasingly using AI to accelerate its own research and development, which could cause model capabilities to advance faster than safety measures are developed and deployed. Anthropic says it has delayed model releases for that reason before, citing Mythos, which was made available on a limited basis through Project Glasswing.
On policy, the filing says the U.S. government could raise costs, restrict operations or introduce rules that affect demand for models. Earlier this year, the U.S. Department of Defense identified Anthropic as a "supply chain risk," and government contract revenue accounts for about 1% of company revenue. The filing concludes that "many of the risks we face are interrelated and may compound or trigger cascading adverse effects."
The seven founders also tell prospective investors not to view a purchase "purely through a trading lens." They say company decisions "may diverge from investor expectations" because profit maximization must be weighed against the "public interest," and some commercial bets "may, in isolation, appear strange or unconventional."
What the market is watching next: October ARR
The filing defines run-rate revenue as fees charged on usage over the most recent 28 days multiplied by 13, plus monthly subscription fees multiplied by 12. The October figure is still blank and is expected to be added in a mid-next-week S-1 update.
In market discussions, more than $73 billion is seen as a reasonable baseline. If the figure exceeds $78 billion, talk of reaching $100 billion by December could pick up again.
Matt Kennedy, senior strategist at Renaissance Capital, told Reuters that no company wants to linger too long once it reaches the edge of going public. It may have been able to secure a higher valuation next year, he said, but Anthropic appears to have decided to push ahead.
One smaller detail in the filing is cultural. Anthropic describes itself as "low-ego," and according to reports, employees internally call themselves "Ants," a nod to the ANT ticker.
