Anthropic, the San Francisco company behind the Claude family of AI models, has finally pulled back the curtain. On Monday it filed its long-awaited S-1 registration statement with the U.S. Securities and Exchange Commission, the formal first step toward what could become one of the largest stock market debuts in history. The company intends to list its Class A shares on the Nasdaq Global Select Market under the ticker symbol “ANT”. Goldman Sachs and Morgan Stanley are named as lead banks, and market watchers expect a listing around Thanksgiving week.
For the first time, investors, rivals and the public can look at real numbers instead of rumors. And the numbers are dramatic in both directions: breathtaking growth on one side, enormous commitments and a headline loss on the other.
The filing does not yet name a share price or the number of shares on offer, which is normal at this stage. Media reports in recent weeks have suggested that Anthropic is aiming for a valuation somewhere between $1.5 trillion and $2 trillion. That range would put a company founded only in 2021 in the same league as SpaceX, which was valued at roughly $1.77 trillion when it went public in June.
Revenue is the part of the story that grabbed the headlines first. In the six months to the end of June 2026, Anthropic booked about $16.3 billion in revenue, compared with roughly $1.16 billion in the same period a year earlier. For the whole of 2025 the company had reported about $4.6 billion, after just $386 million the year before and $41 million in 2023. In the second quarter alone, sales reached around $11.6 billion, and Anthropic reported an operating profit of roughly $974 million for that quarter, its first quarter in the black on that measure.
Most of that money comes from businesses that pay per use, measured in tokens processed by the models. Anthropic calls this consumption-based revenue. On top of that come paid subscription plans for individual users. Coding has been the engine: Claude Code, the company’s programming agent, has grown into a major business of its own, and agent-style products such as Claude Cowork are increasingly sold to companies that started with coding tools and then expanded to other teams.
Compute bills, cloud partners and an unusual power structure
The flip side of the growth is the cost of the machines that make it possible. Computing power is Anthropic’s single biggest expense, and the prospectus does not hide how large the bill could become. In the second quarter of 2026, compute cost the company about $5.6 billion, meaning it earned a little more than two dollars in revenue for every dollar spent on computing. The company says it wants to keep improving that ratio.
Looking forward, the figures become almost abstract. According to the filing, Anthropic has committed to spending between roughly $542 billion and $567 billion on future compute and infrastructure. A large package of deals signed at the end of 2025 alone is worth about $417 billion and runs for up to 20 years. Further capacity starting between 2028 and 2031 adds another $125 billion to $150 billion. Separately, the company agreed a series of deals with SpaceX between May and October 2026 for access to up to $120.5 billion in computing power through 2029.
Much of this capacity comes from Amazon, Google and Microsoft. That makes for a complicated web of relationships. The three tech giants are at the same time Anthropic’s suppliers, its investors, its customers and its competitors. Amazon is reported to have invested about $18 billion and Google about $23 billion. Customers of the three cloud platforms can buy access to Claude through their existing accounts, and that channel accounted for 48 percent of Anthropic’s revenue in the first half of 2026, up from just 11 percent in 2023. The price for that reach is steep: fees paid to channel partners pushed sales, marketing and partnership costs to about $3.2 billion in the first half of the year.
The risk section of the prospectus is unusually frank about what this means. The cloud providers have visibility into Anthropic’s pricing and commercial terms, they might prioritize their own AI products when compute is scarce, and the close ties have already drawn antitrust attention. At the same time, many of Anthropic’s own customers can reduce their spending quickly, while a large share of the compute commitments has to be paid regardless of how much the models are used.
Then there is the bottom line. While the operating loss for the first half of 2026 narrowed to about $693 million, the prospectus shows a net loss of around $80.7 billion for the same period. The gap between those two figures reflects items below the operating line, and analysts will spend the coming weeks digging into what exactly drives it and how much of it is a one-time accounting effect.
Investors will also study the governance model closely, because it is unlike anything typical on Wall Street. Anthropic is a public benefit corporation, which legally obliges it to balance shareholder returns with its stated mission: the responsible development of advanced AI for the long-term benefit of humanity. An independent body called the Long-Term Benefit Trust holds a special class of shares that lets it elect directors who together form a majority of the board. A second special share class, held by an entity owned by the seven co-founders, including CEO Dario Amodei and President Daniela Amodei, will carry 50.1 percent of the voting power on most shareholder matters. Ordinary Class A investors will get one vote per share but, in practice, limited influence.
For some investors this structure is a feature: it signals that the founders cannot easily be pushed into short-term decisions on safety. For others it is a warning sign, because it concentrates control and limits the usual tools shareholders have to correct course.
The document is also unusual in tone. Alongside pages of financial tables, it speaks of a possible coming “age of miracles” driven by AI, while also warning of serious, even existential, risks to society. Few companies have ever asked the public for money while describing their own product in such stark terms.
The timing matters for the whole technology sector. Anthropic will reach the stock market before its arch-rival OpenAI, the company several of its founders left five years ago. A strong debut could confirm that the enormous sums flowing into AI can turn into real, recurring revenue. A weak one could cool enthusiasm at a moment when the industry is committing hundreds of billions of dollars to data centers, chips and power.
Independent research reports suggest that growth has already slowed somewhat in the third quarter, as some large customers trimmed their usage and competition from OpenAI and cheaper open-source models increased. Whether a valuation in the trillions is justified will depend on whether demand for AI agents keeps rising fast enough to fill the vast amount of computing capacity Anthropic has already promised to pay for.
For now, one thing is certain: with this filing, the most closely watched AI company of the year has moved from private speculation to public scrutiny, and every number in it will be tested by the market in the weeks ahead.





