Anthropic files for Nasdaq IPO, disclosing US$80.7bn loss and huge compute commitments

Anthropic, the company behind the Claude models, has lodged a Form S-1 with the US Securities and Exchange Commission for a proposed initial public offering of Class A common stock. The shares are intended to trade on the Nasdaq Global Select Market as "ANT". Neither the number of shares nor the price range has been decided. According to Reuters reporting carried by iTnews, some current and prospective investors now value the company at about US$1.5 trillion, compared with roughly US$2 trillion a week earlier.

First-half 2026 revenue came to about US$16.3 billion. Anthropic said revenue in the three months to 30 June was up 1,368 per cent on a year earlier, while cautioning that the growth rate could fall and swing over time. The net loss for the six months widened to US$80.7 billion, mostly because of a non-operating item linked to bonds that convert into company shares. ChannelNews reported that US$72.4 billion of that was non-cash accounting adjustments, and that the operating loss narrowed to US$693 million.

Anthropic said more than half of Fortune 500 companies and over 30 per cent of Global 2000 firms were customers that each brought in at least US$100,000 over the 12 months to 30 September. ChannelNews reported that its five largest clients made up roughly a quarter of revenue in the first nine months of 2026.

Compute and infrastructure is the biggest cost line. The company said it expects those expenses to keep rising and to stay its largest operating cost. Reuters reported that it has committed more than US$600 billion to computing deals with partners including Amazon and Google. W.Media, citing the prospectus, reported that Anthropic expects to spend as much as US$120.5 billion on Nvidia-powered compute from SpaceXAI into 2029, and as much as US$114.5 billion on Google capacity through 2033. The filing also says the company depends on a small number of cloud providers such as Amazon, Google and Microsoft, and on suppliers such as Broadcom, which can create "dependencies and potential vulnerabilities".

The prospectus flags safety risks, saying its models could show "self-preserving behaviours" and behaviour "resembling blackmail". It also says using AI in its own research may push model capabilities ahead of its safety work. Goldman Sachs has the lead-left role alongside Morgan Stanley. Reuters reported the listing is expected before late November, while W.Media said it reportedly targets the week of 16 November.

Analysts see the offering as a test for the sector. IPOX research associate Lukas Muehlbauer said the IPO "will establish a benchmark for valuing leading AI labs, influencing both upcoming listings and confidence in the companies funding the AI infrastructure build-out". deVere Group chief executive Nigel Green was more critical, saying: "Frankly, it's nuts,".