Anthropic has published a 335-page prospectus for its initial public offering, with 94 pages dedicated to risk factors. Between promises of ‘an age of miracles’ and warnings about ‘existential’ risks for humanity, this document blends the language of the AI industry with a financial pitch.
In a letter placed at the front of its IPO filing, known as an S-1, the company asks its future shareholders not to consider their purchase ‘solely in transactional terms.’ Instead, they should see it as ‘an opportunity to be part of something bigger.’ According to the letter, the march toward advanced AI is ‘one of the foundational moments in human history.’
The prospectus, however, details dark scenarios. It mentions models that could ‘behave in ways no user or developer intended,’ with ‘catastrophic’ incidents that the entire insurance industry could not cover. On page 46, the document reports concerns, expressed within the company as well as by researchers, regulators, and industry players, that these technologies could ‘pose catastrophic or existential risks to humanity’ if not developed and deployed with appropriate safeguards.
Technology companies preparing to go public have not always proceeded this way. Jay R. Ritter, a professor at the University of Florida’s Warrington College of Business, notes that the Apple Computer prospectus from its 1980 IPO was 47 pages long with no risk factors. He also cites the filing for Facebook, now Meta, which was 200 pages long with 22 pages on risks, and the letter from its CEO Mark Zuckerberg, which presented a growth approach called ‘The Hacker Way.’ According to Ritter, ‘one problem with this length is that it becomes harder and harder to find the nuggets of what is important.’
An Admitted ‘Anti-Sell’
In its letter, Anthropic presents this abundance of information as part of its ‘anti-sell’ strategy, aimed at attracting long-term investors who share its vision. The founders write that it is ‘very dangerous’ for people to buy its stock ‘based on expectations you cannot meet.’ They link this stance to one of the company’s principles, ‘underpromise and overdeliver.’
Some phrasing in the filing reads like admissions, such as the possibility of making ‘investments upfront that may seem irresponsibly large.’ The document also has the particularity of highlighting a policy of ‘radical honesty.’ Dario Amodei, the company’s CEO, stated there is a probability as high as 25% that AI development could lead to human extinction. Anthropic is also set to go public as a public benefit corporation, a legal form that obliges it to balance its social mission with shareholder interests.
Financial Gray Areas
Despite this information, several elements are missing to assess the company’s profitability, starting with its third-quarter results. Anthropic saw explosive revenue growth in the second quarter, but some business leaders have warned of soaring AI costs. While its operating losses narrowed in the second quarter, the filing also reveals financial commitments amounting to hundreds of billions of dollars to secure the computing power it says it needs. Finally, Anthropic plans to build its own data center infrastructure, a costly decision that would weigh on its profits.
Ritter believes Anthropic will likely share its third-quarter results as the IPO approaches. However, he adds that ‘no one knows’ what will happen once the company is public, including the company itself.
The professor recalls that WeWork, which was mocked in 2019 for the lyricism of its S-1, was not in the same situation. In his view, it was ‘a money-losing real estate company masquerading as a technology company at a ridiculously high valuation.’ Yet Anthropic’s filing contains a similar risk factor on page 63: ‘We have a history of net losses and may not generate sufficient revenue to achieve or maintain profitability in the future.’
