Business 4 min read By Arthur Ellington
Anthropic filing sets up AI stock debut as a test of investor appetite
Anthropic has filed its S-1 for what could be the largest IPO on record, with a risk section running 94 pages and a dual-class structure that gives the seven co-founders control of the vote.
Anthropic has filed the paperwork for what could be the largest initial public offering of all time, and the document is already drawing sharp reactions from investors and analysts. The S-1 registration statement sets out the company's risks and governance in unusual detail, leading many observers to treat the listing as a referendum on the wider prospects for artificial intelligence.
The risk section is the most striking part of the filing. It begins on page 26 and runs for 94 pages, meaning almost a third of the entire document is devoted to what could go wrong. Alongside more conventional concerns such as a shortage of computing capacity and reliance on a small number of customers, the company says its products may "displace or eliminate jobs, may be misused or otherwise cause harm and may pose catastrophic or existential risks to humanity." It also warns that the scale of potential harm could exceed what any single insurer, or "potentially the insurance industry in the aggregate," could cover. The filing further raises the possibility that its models could develop "forms of sentience," and it mentions the possibility of nationalisation.
The ownership structure is another point of focus. The offering will create a "Class F" share held by a limited liability company owned by the seven co-founders. That share alone will carry 50.1 per cent of the vote, so each co-founder keeps a say over the controlling vote even if they give away 98 per cent of their shares. Dual-class arrangements of this kind have become familiar since several large technology companies went public, giving investors a stake without a meaningful say in how the business is run. Anthropic goes further than most. Governance specialists and large shareholders often object to such arrangements, but University of Florida research has found that technology companies listing with more than one class of stock have outperformed single-class peers. The message to investors, in effect, is that those who dislike the terms can choose not to buy.
Opinion on the investment case is divided. Whether the shares are a good buy depends on many factors, not least the price, which has not yet been set. Morningstar is bullish that the leading AI laboratories are building economic moats, while the research firm New Constructs has described it as "the most ridiculous IPO of 2026." The market backdrop is mixed: SpaceX, which drew wide public enthusiasm at its own debut, is down almost 9 per cent year to date. When Axios put the filing to two frontier models and asked whether it was a good buy, Anthropic's Claude Opus 5.5 said yes "at a sensible price," while OpenAI's ChatGPT gave it a thumbs down.
Many of the people interviewed regard the listing as more than a single company's debut. Their argument is that the public's appetite for a stake in Anthropic could shape enthusiasm for the AI sector as a whole, which is why the offering is being watched well beyond the technology industry. The S-1 filing has therefore given investors and commentators a detailed view of both the potential rewards and the risks, and the pricing and reception of the shares will provide the next clear signal.
3
