Anthropic's IPO puts its own warning on entry-level jobs in front of investors and graduates

Anthropic is heading to the public markets. It filed its S-1 prospectus, the document that sets out a company's business and risks for investors, on October 26. It plans to list on the Nasdaq under the ticker ANT. Fortune reports the listing could come around Thanksgiving, and Reuters reports investors value the company at $1.5 trillion to $2 trillion. That would make it one of the largest IPOs in history. The awkward backdrop is that in a May 2025 interview, CEO Dario Amodei warned that AI could more than double the unemployment rate within one to five years.

Forecasters are starting to put numbers on that warning. Marcos Ortega, a professional forecaster on the Metaculus platform, told the Independent Florida Alligator he expects US unemployment to reach roughly 8% to 10% within about five years. That is about double the current 4.2% reported by the Bureau of Labor Statistics. He does not expect mass layoffs. He expects a 'no firing, no hiring' pattern instead: companies keep experienced staff who know the internal workflows and stop opening junior roles. In his view, most of the losses will come from knowledge work done at a computer. He sees finance and computer science graduates as especially exposed, because their output is easy to check and therefore easier to train AI on. Metaculus forecasters expect US employment of financial specialists to fall about 13% between 2025 and 2035. Registered nurses are one of the few roles they expect to grow.

The local numbers show the mismatch. At the University of Florida, about 2,100 of the 10,160 bachelor's degrees awarded in 2024-25 went to business or computer science. Only about 730, or 7%, went to health professions.

Governance is the other issue the paper raises. According to the prospectus, Anthropic's seven co-founders will hold 50.1% of the votes on major shareholder decisions through a single special share. That control only starts to phase out when two or fewer founders remain. An independent Long-Term Benefit Trust would fill four of the seven board seats. Its trustees serve one-year terms and elect their own successors. MIT researcher Marco Mari points out that founder control is common: Mark Zuckerberg holds 60.8% of Meta's voting power. He also notes that Facebook's social costs, such as Cambridge Analytica, surfaced six years after its IPO. With AI, he argues, the effects on labor and science are visible before the company even lists.

Our reading: the most important point here is not the size of the valuation. It is the 'no hiring' mechanism. Job losses from AI may not show up as mass layoffs. They may show up as a missing first rung on the career ladder. That fits our long-running thesis: the most exposed work is routine and easily verified, while judgment, relationships and in-person care hold up. The forecasts are forecasts, not data, and the 8-10% figure comes from one forecaster. Still, they point the same way as Amodei's own warning. Going public will not speed up automation, but it will bring more scrutiny. We think that is welcome for a company whose main product could reshape the labor market. In the short term, junior workers carry the cost. Over the long term, the abundance AI promises only becomes real if new ways into work are built to replace the ones being closed. That job belongs to universities, employers and policymakers, not to Anthropic's founders alone.