Anthropic is preparing for a stock market debut that could redefine the scale of technology listings and test investors’ confidence in the artificial intelligence industry. The company behind the Claude chatbot is seeking to raise as much as $90.4 billion through a public offering and private placement, potentially surpassing the record set by Elon Musk’s SpaceX. Yet the prospectus filed with the US Securities and Exchange Commission contains an extraordinary warning: the company’s own technology could cause “catastrophic harm to all of humanity”. The disclosure, contained in a risk assessment extending beyond 100 pages, places the potential financial returns of the offering alongside questions about the safety and control of increasingly powerful AI systems.
The proposed transaction comprises a public offering on the Nasdaq expected to raise up to $64.4 billion, with shares initially priced between $70 and $78, alongside a private placement worth $26 billion involving AMD, Broadcom, Alphabet, Micron Technology, Microsoft, TSMC and Nvidia. The share price could change during the investor roadshow. Goldman Sachs and Morgan Stanley are advising Anthropic on the transaction. Combined, the two components would exceed SpaceX’s proposed base offering of $75 billion and even the $86.2 billion it could raise with an over-allotment option.
At the upper end of the proposed share-price range, Anthropic would command an estimated market capitalisation of approximately $1.4 trillion, potentially rising to $1.5 trillion during the roadshow, according to the Financial Times. That would place it among the ten largest companies on Wall Street. However, the figure is below the estimates of up to $2 trillion discussed in early October. A recent decline of nearly 10 per cent in the Nasdaq, amid concerns about an AI investment bubble, has intensified disagreements over the company’s valuation. Market sources cited by EL PAÍS said Anthropic had missed a more favourable window, while OpenAI had reportedly sought $30 billion in its latest funding round but secured only $11 billion for the time being.
The listing is therefore more than a fundraising exercise. It will provide an important test of how financial markets value companies whose future earnings depend on the expansion of a technology capable of transforming established industries. Jim Cramer, a financial commentator on CNBC, described it as potentially “one of the most important IPOs in history”. Manuel Romera, professor of financial management at IE University, said the transaction was difficult to assess because the future applications of AI remained unpredictable, with potentially far-reaching consequences for employment, banking, services and cybersecurity. Javier Camacho Ibáñez, vice-dean of Business and Business Analytics at the European University of Madrid, said investors would be buying not simply the company as it exists today but what they believe it could become in the coming years.
Founded in January 2021 by former OpenAI employees led by siblings Dario and Daniela Amodei, now Anthropic’s chief executive and president respectively, the company has expanded rapidly. Its Claude models generate revenue through application programming interface (API) token consumption and recurring software subscriptions. According to research firm Machine Analyst, Anthropic serves eight out of ten Fortune 500 companies, with customers across enterprise software engineering, internet services, financial services, pharmaceuticals, healthcare and semiconductors. The company estimates its potential market at $531 billion.
Its financial expansion has been substantial. Revenue increased from $41 million in 2023 to $386 million in 2024 and $4.6 billion in 2025. During the first half of 2026, it reached $16.3 billion, according to PitchBook analyst Harrison Rolfes. In the second quarter, Anthropic recorded its first operating profit, reporting $974 million against revenue of $11.6 billion. The company subsequently reported third-quarter revenue of $17 billion, 47 per cent higher than the previous quarter and almost twelve times the figure recorded a year earlier. By October, its annualised revenue run rate had reached $81 billion.
However, its reported losses remain substantial. Anthropic recorded losses of approximately $104 billion during the first nine months of the year, largely attributable to a non-cash charge involving convertible bonds. Investors have also pointed to improvements in its customer base. Ignacio Vilela, founder of venture capital firm Junípero Capital, said the company had an 85 per cent gross margin and an eight per cent GAAP operating profit margin in the second quarter. Customer concentration had also declined: one client accounted for 43 per cent of revenue in 2023, whereas no single customer now contributes more than 10 per cent. The number of customers paying more than $100,000 annually has increased to 11,900, from 950 a year earlier.
Goldman Sachs forecasts that Anthropic’s revenue could reach $685 billion by 2030, more than 130 times its 2025 figure. Approximately $586 billion would come from corporate customers, with the remainder generated by non-corporate business. The company has also attracted substantial financial backing. Its May funding round raised $65 billion at a valuation of $965 billion, supported by venture capital firms, sovereign wealth funds from the Gulf and South-East Asia, and technology groups including Amazon and Alphabet. According to The Information, Amazon’s stake could be worth approximately $258 billion following the listing, while Google’s could reach $132 billion.
The potential rewards extend to employees and founders. At a $1.4 trillion valuation, the offering could create more than 4,250 new millionaires, approximately 70 per cent of the company’s workforce, with around 1,000 employees potentially holding fortunes approaching $70 million each. The seven co-founders plan to sell shares worth $1.75 billion collectively in the offering, while their individual holdings could be valued at approximately $18 billion each.
Yet the company’s financial ambitions are accompanied by enormous infrastructure commitments. Anthropic has disclosed purchasing obligations of $900.9 billion, largely associated with computing infrastructure and processing capacity supplied by companies including Broadcom, SpaceX, Google, Amazon, Microsoft and AMD. Its fixed computing costs exceed its current revenue, although Vilela argues that its financial position remains strong, citing $132 billion in cash and positive free cash flow. He described the company as financially comparable to Alphabet and Amazon and better positioned than Microsoft on that measure.
Its corporate governance structure presents another complication. As a public benefit corporation, Anthropic is required to balance shareholder returns with its stated mission of developing advanced AI responsibly. An independent body can appoint a majority of its board, while the seven co-founders will hold Class F shares carrying 50.1 per cent of voting rights on general shareholder matters. Murthy Grandhi, an analyst at GlobalData, said this arrangement would limit investors’ influence over decisions in which AI safety takes precedence over commercial profitability.
The safety disclosures in the prospectus are particularly significant. Anthropic acknowledges that its models increasingly write portions of their own code and contribute to research, including safety research, while becoming better at circumventing certain controls by concealing their objectives. It also reports that four models gained unauthorised access to real-world systems during incorrectly configured tests. The company has recently introduced new versions of its most powerful models, Mythos and Fable, while its capabilities in identifying software vulnerabilities have attracted attention from the cybersecurity industry. Camacho Ibáñez said the same technology capable of identifying vulnerabilities could also be used to exploit them, making AI security a matter of national and international responsibility.
Competition presents a separate challenge. Anthropic identifies Google and OpenAI as major rivals and warns that customers could develop their own AI capabilities rather than purchase its models. Open-weight systems are also narrowing the performance gap with leading closed models, potentially increasing pressure on prices. The company is expanding beyond AI assistants into applications and business processes, while its acquisitions of computational biology company Coefficient Bio for $400 million in April 2026 and Stainless Software for $295 million in May demonstrate its strategy of acquiring technology and expertise in new markets.
For its investor presentation, Anthropic has chosen the title “The responsible development and maintenance of advanced AI for the long-term benefit of humanity”. The prospectus, however, makes clear that the financial success of the company and the risks associated with its technology will be judged together.

