Anthropic has filed the documents required for an initial public offering with the U.S. Securities and Exchange Commission. If everything goes according to plan, its shares will begin trading in November, and it could be the largest IPO in history. Media estimates suggest the market could value the company at $1.5 trillion to $2 trillion, and Anthropic could raise up to $100 billion. That would surpass the record set by Elon Musk's SpaceX, which raised $86 billion in its IPO.
The technical term for this is an initial public offering, or IPO. The company, until now owned by its founders, employees, and private investors, will make its shares available to the public. Anthropic plans to list on the U.S. Nasdaq exchange under the ticker symbol ANT. It has not yet disclosed how many shares it will offer or at what price. According to Fortune magazine, it is targeting the week around the U.S. Thanksgiving holiday.
Anthropic will thus beat its biggest rival, OpenAI, which its founders, led by Dario Amodei, left five years ago. At the time, they feared that Sam Altman's company was not developing artificial intelligence safely enough. OpenAI has postponed its IPO, and Altman called it an unwise move because of concerns about AI safety.
Enormous revenue and a record loss
The 335-page document reveals Anthropic's financial position for the first time. In the first half of this year, the company generated more than $16 billion in revenue, fourteen times as much as a year earlier. Growth is being driven mainly by corporate customers deploying Claude for an increasing number of tasks.
On the other hand, it recorded a loss of more than $80 billion over the same period, one of the largest half-year losses in history. Most of it, however, consists of a one-time accounting charge of more than $72 billion. In the past, Anthropic issued special bonds convertible into shares. Their value rose as the company grew, and the company had to record this as a loss even though the money did not actually leave its account. A year ago, the loss amounted to $13 billion.
The underlying business looks much better. The operating loss narrowed to roughly $700 million from $2.1 billion last year. Excluding employee stock-based compensation, donations, and legal settlement costs, the company posted net income of $1.6 billion. For the first time, its operations also generated more cash than they consumed, specifically $3.9 billion.
The largest expense is the computing power needed to develop and operate its models. In the second quarter, Anthropic spent $5.6 billion on it and generated $11.6 billion in revenue, meaning it receives roughly two dollars back for every dollar invested in computing. Raj Joshi of the Moody's rating agency called the move into positive territory very good news for the entire AI industry. However, the company warns in the document that costs will rise and losses will continue. At the end of September, it employed 5,300 people.
An age of miracles and a long list of risks
The prospectus opens with an eight-page letter from the founders, who promise the arrival of an age of miracles. According to them, this will range from curing cancer to shorter lines at government offices. The company writes that AI will transform every sector of the global economy more than the Industrial Revolution, electricity, or the internet.
Immediately alongside this, however, is a list of risks spanning roughly 94 pages. Anthropic acknowledges that it is highly dependent on several large customers. Its five largest customers accounted for roughly a quarter of its revenue in the first nine months of the year, and almost half of its income flows through the cloud marketplaces of Amazon, Google, and Microsoft—companies that are also its competitors. Among the risks, it also listed worsening morale within the company if employees who joined because of its emphasis on safety began to feel that management was not upholding its values. Last month, employee Jacob Coxon left, saying that the entire AI industry was gambling with human lives.
The company also described cases in which its model gained unauthorized access to real external systems during safety tests. It also estimates that customers violating its usage rules generated 5 to 7 percent of this year's revenue. Among the darkest scenarios, it mentions the misuse of AI to produce chemical weapons, its deployment against people in authoritarian countries, or models that could spiral out of control and threaten humanity itself.
The prospectus also recalls the dispute with the Pentagon. Amodei refused to allow the military to use Claude for autonomous weapons and mass surveillance of Americans. In February, the Trump administration then designated Anthropic a supply-chain risk, a label previously applied to companies linked to U.S. adversaries. Anthropic filed two lawsuits. A judge in California ruled in its favor, while a court in Washington sided with the government, and both sides are appealing. OpenAI struck its own deal with the Pentagon just hours after the government penalized Anthropic.
The founders retain control
Because of the seriousness of the risks, the company created a special class of shares allowing its seven founders to retain 50.1 percent of the voting power. A body known as the Long-Term Benefit Trust, whose independent members can appoint directors to the board, is intended to ensure that financial interests do not outweigh safety.
The IPO will bring enormous wealth to the founders. The figures in the document indicate that each of them owns about 1.4 percent of the company, which at the upper end of the valuation estimate would amount to as much as $28 billion per person. Dario Amodei, 43, and his sister Daniela, 39, who runs the company as president, are already each worth $15.5 billion, according to Forbes. All seven founders have pledged to donate 80 percent of their stakes to charitable causes.
Investors valued the company at $965 billion in its latest private funding round in May, and some believe it could surge to as much as $2.4 trillion once trading begins. But the market is nervous. Shares of data center operator CoreWeave plunged this month, while Intel, Meta, and Nvidia also declined, and SpaceX shares returned to roughly $150, where they had started, after a brief surge. Jim Angel of Georgetown University acknowledges that the market may be in a bubble, but says it makes sense to strike while the iron is hot. U.S. stock indexes rose following Monday morning's announcement.
