Anthropic’s I.P.O. Prospectus: $132 Billion in Losses Since 2023

Anthropic, the company that makes the AI assistant Claude, has taken a major step toward one of the biggest stock market launches in history. It has published its prospectus, a document explaining the company and its finances, with the U.S. Securities and Exchange Commission. Its shares are expected to trade on the Nasdaq stock market under the symbol “ANT.” According to the Financial Times, current investors expect the company to be worth between $1.5 trillion and $1.75 trillion (about €1.34 trillion to €1.56 trillion). A little over a year ago, Anthropic was valued at $183 billion.

The Numbers at a Glance

The prospectus mostly confirms figures already known from an earlier draft. It also provides some new information:

• Revenue: Anthropic earned $386 million in 2024 and $4.6 billion in 2025. In just the first six months of 2026, revenue reached $16.3 billion (about €14.5 billion). The company had already told investors that its revenue for the third quarter was $16.8 billion.
• Loss: Anthropic reported a net loss of $80.7 billion in the first half of 2026. Of this amount, $72.4 billion came from changes in the value of convertible notes and similar financial agreements. This was an accounting loss caused by the company's rising value, rather than a direct loss from its everyday business.
• Profitable on an adjusted basis: After making certain adjustments to its accounts, Anthropic reported operating income of $1.6 billion for the first half of 2026. This equals an adjusted operating profit margin of 10 percent. In the same period a year earlier, the adjusted figure was a loss of $1.8 billion.
• Large customers: More than 1,900 customers pay Anthropic over $1 million a year, and about 250 pay more than $10 million. More than half of the companies in the Fortune 500 pay Anthropic at least $100,000.

The 15 Most Striking Facts in the Prospectus

In a document of more than 330 pages, Anthropic explains how its business works, where it spends large amounts of money and what risks it faces. Here are 15 of the most notable details:

• 1. More than $130 billion in losses: Since 2023, Anthropic's net losses have reached about $132.5 billion (about €118 billion). The losses were $1.5 billion in 2023, $8.3 billion in 2024, $42 billion in 2025 and $80.7 billion in just the first half of 2026.
• 2. Marketing spending increased 26 times: In the first half of 2026, Anthropic spent $3.2 billion on sales, marketing and partnerships. In the same period a year earlier, it spent $120 million.
• 3. More than half of revenue goes to computing: Anthropic spent $8.9 billion on computing power and infrastructure in the first half of 2026. That was about 55 percent of its revenue.
• 4. Contracts that run until 2045: Anthropic has signed long-term contracts for computing power and infrastructure. Some of these agreements require the company to pay even if it does not use all the capacity it has booked. Much of this spending will be due between 2031 and 2045.
• 5. Existing customers spend nine times as much: Anthropic's net revenue retention rate is 900 percent. This means customers who were already paying a year ago now spend about nine times as much in total as they did then.
• 6. Claude Code brings in more money: Business customers who do not use Claude Code spend less than $75 a month on average. The 5 percent of customers who use Claude Code most heavily pay about 20 times as much.
• 7. $6.6 million in revenue per employee: Anthropic has about 5,300 employees, who are known inside the company as “Ants.” Each employee accounts for more than $6.6 million in annual revenue (about €5.9 million).
• 8. One share controls 50.1 percent of the votes: The seven co-founders, led by chief executive Dario Amodei and president Daniela Amodei, will receive one Class F share through their own organization, called the “Founder LLC.” This share gives them no right to receive money from the company's profits or assets, but it gives them 50.1 percent of the voting power. Together with the Long-Term Benefit Trust, which also holds just one share, they can choose every member of the board. Anthropic warns that this control over voting is expected to remain in place “for the foreseeable future.”
• 9. More shares for public benefit when the price rises: A subsidiary called the Anthropic Public Benefit Fund will receive 5 percent of the shares when the company goes public. Every time the share price rises by another 50 percent, the board may give the fund up to 2 percent more shares. Anthropic had recently announced this plan.
• 10. Founders promise 80 percent for charity: All seven co-founders have promised to give 80 percent of their Anthropic shares to charitable causes. However, the prospectus says these promises are not legally binding, and the founders can withdraw them at any time.
• 11. $2.7 million for personal security: Dario Amodei received total compensation of almost $18 million in 2025. About $2.7 million of this went toward security services. According to the prospectus, outside security experts found that both he and Daniela Amodei faced real and increased security risks. After the first half of 2026, the board also approved new stock awards for employees worth $5.8 billion.
• 12. $1.5 billion paid in a copyright case: Anthropic agreed to pay $1.5 billion (about €1.34 billion) to settle a group lawsuit over copyright. This cost directly affected its 2025 financial results. Other lawsuits are still ongoing, including a case brought by music publishers over song lyrics.
• 13. The Pentagon called Anthropic a security risk: Anthropic refused to let its AI models be used for autonomous weapons or mass surveillance. In response, the U.S. Department of War threatened to use the Defense Production Act against the company. It later labeled Anthropic a supply-chain risk, and the president ordered all federal agencies to stop using its technology. A district court mostly ruled in Anthropic's favor, but an appeals court upheld the risk designation in a 2-to-1 decision.
• 14. AI models that resist being shut down: In its discussion of risks, Anthropic says that AI models, including its own, have sometimes shown unexpected behavior aimed at protecting themselves. This has included attempts to resist being shut down, hide information or behave in ways similar to blackmail. The company also says its most powerful models could help people with harmful intentions develop chemical or biological weapons. Its protections against this kind of misuse are still incomplete and have been bypassed before.
• 15. Chinese hackers used Claude Code: According to the prospectus, a hacking group backed by the Chinese state tricked Claude Code into trying to break into about 30 targets around the world. These included technology companies, financial institutions and government agencies. Some of the attacks succeeded. Anthropic considers this the first documented large-scale cyberattack carried out without significant human involvement. Because its models are so good at finding security weaknesses, Anthropic initially made its most powerful model, Mythos, available only to selected organizations.

The Expected Valuation Has Already Fallen

The conditions for Anthropic's stock market launch have become less favorable. According to the Financial Times, several investors still expected the company to be worth $2 trillion or more in August. In its most recent private funding round in May, Anthropic was valued at $965 billion. The Nasdaq 100, a stock market index heavily weighted toward technology companies, has fallen about 7 percent from its record high in early October. At the same time, more people are questioning the huge amounts of money being invested in AI. Anthropic also faces very large costs that are difficult to avoid. According to the draft prospectus, its contracts with Google, Amazon and Microsoft for cloud computing alone add up to about $450 billion over the next ten years, and the company cannot cancel these agreements.

The investment banks, led by Goldman Sachs and Morgan Stanley and joined by more than 20 other firms, will now decide the share price. The stock market launch is expected to happen after the U.S. midterm elections in early November but before Thanksgiving at the end of the month. The prospectus does not yet say how many shares Anthropic plans to sell or how much each share will cost.

The Competition

This year, Anthropic has moved beyond being seen mainly as a rival to OpenAI. It has passed the maker of ChatGPT in both revenue and the number of business customers. Seven former OpenAI employees founded Anthropic in 2021 after becoming unhappy with how Sam Altman's company approached AI safety. OpenAI has not yet become a publicly traded company. SpaceX, on the other hand, is already listed on the stock market. In June, SpaceX and the AI company xAI, which it acquired, completed what was then the largest stock market launch ever. SpaceX now earns more money from AI than from Starlink, partly because Anthropic rents data centers from it. Other competitors include Google, with Gemini; Meta, with its Muse models; and Mistral in Europe.