Anthropic has filed its S-1 registration statement with the US Securities and Exchange Commission (SEC) for an initial public offering (IPO) of Class A shares. It will list on the Nasdaq Global Select Market under the symbol “ANT”. The filing does not yet disclose the price band or the number of shares on offer.
Goldman Sachs and Morgan Stanley are leading a group of 22 banks. The issue will include newly issued shares as well as shares sold by existing investors. Anthropic will not receive any proceeds from the sale of shares by existing investors. Besides the US, the shares will be offered to investors in Australia, Canada, parts of Europe, Japan, Switzerland and the UK. Anthropic will list only on Nasdaq.
According to Bloomberg, marketing is scheduled to begin on November 9, with the company targeting a listing in the week of November 16. Media reports suggest the IPO could raise around $100 billion at a valuation of $1.5 trillion to $2 trillion. Anthropic was valued at $965 billion in a private funding round in May.
Here are six things that stand out in the filing.
1. Revenue grew 14 times in a year
Anthropic generated $41 million in revenue in 2023, $386 million in 2024 and $4,585 million in 2025. In the first six months of 2026, revenue reached $16,286 million, 14 times the $1,163 million recorded in the same period last year.
Customers are also spending more each year. Anthropic’s net revenue retention is about 900%, meaning existing customers are paying roughly nine times what they paid a year earlier. As of September 30, it had more than 11,900 customers paying at least $100,000 a year, more than 1,900 paying at least $1 million and about 250 paying at least $10 million.
The company attributes much of this growth to Claude Code, its coding agent. It says Claude Cowork, its tool for general office work launched in January 2026, is being adopted faster than Claude Code was at the same stage.
2. A huge loss on paper, but a profit on an adjusted basis
Anthropic reported a net loss of $80,748 million for January to June 2026. Of this, $72,388 million came from an increase in the value of its convertible notes. These are loans that will convert into shares. As Anthropic’s valuation rose, the value of the notes on its books increased. This is an accounting entry rather than cash spent. The same item accounted for $34,308 million of its $41,965 million loss in 2025.
The business itself is close to breaking even. The operating loss for the first half was $693 million, compared with $8,060 million for the whole of 2025. On an adjusted basis, Anthropic made an operating profit of $1,618 million in the first half, representing a 10% margin. A year earlier, its adjusted operating margin was minus 159%.
Anthropic has raised $141.1 billion in total so far. As of June 30, it had $112.4 billion in cash and $19.3 billion in short-term investments.
Also read: Anthropic files for IPO, moves closer to becoming first frontier AI lab to go public
3. Spending on compute is huge, but growing more slowly than revenue
Computing power, including the chips and data centres used to train and run its models, is Anthropic’s biggest cost. It spent $8,945 million on compute and infrastructure in the first half of 2026, more than the $7,326 million it spent in the whole of 2025.
However, revenue is growing much faster. In the April to June quarter, revenue rose 1,368% from a year earlier, while compute costs increased 300%. Anthropic also says its newest model, Claude Opus 5.5, reduced the cost per task for customers by about 40% compared with Claude Opus 5.
The filing confirms that the company has signed long-term compute agreements, most of which run between 2031 and 2045. The total value has been left blank for now and will be filled in closer to the listing. Anthropic runs its models on chips from Amazon, Google, NVIDIA and AMD, and plans to build more of its own data centres.
Sales and marketing costs also rose sharply, from $120 million in the first half of 2025 to $3,168 million in the same period this year. The 2025 accounts include a one-off $1.5 billion charge for the Bartz settlement, a copyright case brought by authors.
4. Its cloud partners are also its competitors
Anthropic sells Claude through Amazon Web Services, Google Cloud and Microsoft Azure, while buying most of its computing power from the same companies. These companies also develop their own AI models.
The filing says this conflict is real. Anthropic says it has asked these cloud providers for more computing capacity than they were willing to provide. It warns that its access to compute could be reduced, repriced or terminated, whether for competitive reasons or otherwise. It also says that a large part of its revenue depends on a small number of customers and sales channels.
5. Risk factors few other IPOs have carried
Alongside the usual business risks, Anthropic has included risks relating to its own technology in the filing. It says concerns have been raised, including within the company, that advanced AI systems could pose “catastrophic or existential risks to humanity” if they are not developed and operated with appropriate safeguards. It says this risk is higher if such systems are misused, copied, turned into weapons or allowed to act with a high degree of independence.
The filing also warns that its models may behave in unintended or harmful ways that no user has requested, and that this risk increases as models become more capable and are given greater access to other systems. It adds that even a single instance of such behaviour, whether during testing or in real-world use, could trigger public concern, regulatory action and lawsuits, even if no actual harm occurs. It says its insurance and contracts may not fully cover the damage if its AI causes harm.
Anthropic also warns that, because Claude is now helping to build the next versions of Claude, its models could improve faster than its safety measures can keep pace. It says it has delayed or restricted model releases in the past for this reason, citing its Mythos model, which it made available only to select partners through Project Glasswing. It says such steps could put it at a disadvantage against rivals that do not take similar precautions.
6. The Founder LLC: one share, 50.1% of the vote
Just before the IPO, Anthropic will issue one share of a special Class F stock to a company set up by its seven co-founders, called the Founder LLC. The founders are Dario Amodei, Daniela Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish and Christopher Olah. The filing describes them as “distinctly equipped to be stewards of our mission”.
That single share carries 50.1% of the total voting power on matters put to shareholders, including the election of directors not chosen by the trust. It has no economic rights, cannot be sold and can be cancelled by the company in certain circumstances. Reuters, which saw an earlier draft, reported that a founder can be removed from the LLC for leaving, dying, selling too many shares or for cause. It also reported that the special share starts to phase out once only two or fewer founders remain.
Also read: SpaceX secures $40 billion lease to expand AI computing capacity with Nvidia chips
The founders do not control the board alone. Anthropic’s Long-Term Benefit Trust, a body of five trustees with no financial stake in the company, will elect a majority of the board through a separate Class T share. Together, the trust and the Founder LLC can elect every director. Because the trust controls a majority of the board, Anthropic qualifies as a “controlled company” under Nasdaq rules.
Public investors will buy Class A shares, each carrying one vote. Strategic partners hold Class N shares, which carry no votes and convert into Class A shares when sold. Employees will receive Class E shares, which can vote only on a limited number of governance matters. The filing acknowledges that this structure could leave Class A shareholders with little say over important decisions, including a sale of the company.
As a public benefit corporation, Anthropic is legally required to balance shareholder returns with the interests of people affected by its work and its goal of developing AI responsibly. It has allocated 5% of the company to an Anthropic Public Benefit Fund. The board can add up to 2% more each time the share price rises 50% above the previous milestone, starting from the IPO price. Each co-founder has also pledged 80% of their Anthropic shares to charity.
What comes next
The next filing will add the price band, issue size and total value of compute commitments. That figure, along with the price, will help investors assess a company growing faster than almost any before it, but controlled by its founders and a trust.