The Doomers Are Headed to Wall Street

On the day Anthropic filed its S-1 for a US IPO — Reuters confirmed the filing and Business Insider reported the same day that OpenAI and Anthropic were executing a coordinated pivot on AI job displacement — the company's chief executive had spent the previous twelve months predicting that AI would eliminate half of all entry-level white-collar jobs and send unemployment to 10 or even 20 percent. The empirical record, meanwhile, had produced a different verdict.

Yale Budget Lab, running actual employment figures through March 2026, found no measurable AI-driven displacement in high-exposure occupations. Challenger, Gray & Christmas, the outplacement firm, counted 49,135 tech jobs cut and cited as AI-related in the first four months of 2026 — 26 percent of all layoffs announced in April. Tech companies broadly have now passed 115,000 cuts on the year, a figure that includes large AI developers alongside the broader sector. Whether OpenAI or Anthropic specifically are among the AI-cited cuts remains unconfirmed by Challenger Gray's count, but the broader pattern holds: the industry's most prominent safety voices are also reducing headcount.

Dario Amodei has heard the data. He updated accordingly, adopting a new intellectual framework: Jevons Paradox, the economic principle that efficiency gains expand rather than contract demand. "If you automate 90 percent of the job, then everyone does the 10 percent of the job," Amodei told a JPMorgan Chase financial briefing, "and the 10 percent kind of expands to be 100 percent of what people do." Sam Altman took a different path — telling TIME he was delighted to be wrong about AI job impact and that entry-level white-collar jobs have not been eliminated as he predicted. Both conclusions point away from the catastrophe both men described a year ago. The intellectual routes diverged from there.

The contradiction is worth dwelling on. Amodei has also warned, repeatedly, that AI is moving faster than all previous technologies and that the rebalancing may not arrive fast enough to spare workers displaced in the meantime. The Jevons mechanism, if it works as he describes, eliminates the need for that warning — efficiency gains expand demand, disruption is smoothed, workers are absorbed. If Jevons resolves displacement via demand expansion, the caveat that rebalancing may not come fast enough is either unfounded or refers to a transition period Amodei has not quantified. He has offered no such number. Holding both claims — Jevons as the investor frame, the rebalancing caveat as the public record — requires an explanation he has not given. The investors hearing the Jevons pitch are the same bankers who will underwrite his company's public listing. The workers hearing the rebalancing caveat have no such certainty about when, or whether, it resolves.

The pivot's timing is harder to explain than the pivot itself. Anthropic has filed its S-1. OpenAI, meanwhile, has raised tens of billions in private capital and has signaled listing ambitions. Both companies are moving toward public markets where quarterly earnings calls will force direct answers on headcount, margins, and competitive positioning. The safety-first narrative becomes structurally harder to maintain under that kind of disclosure pressure.

Altman has updated his position as well. He tried delegating his Slack and email to AI, then returned to manual response because, he said, he values human interaction. Goldman Sachs chief executive David Solomon, citing the bank's own research, pointed to US civilian employment growth of 145 percent since 1962 and data center construction adding 200,000 jobs since 2022.

The unemployment data has not contradicted them. The rate has ticked up slightly since April 2024, from 3.9 to 4.3 percent, with no evidence of the displacement the CEOs predicted. The public has noticed anyway. An NBC poll in March found AI with the third-lowest net positive rating of any topic surveyed, above only the Democratic party and Iran. Gallup polling found Gen Z increasingly anxious and angry about AI. A man was charged in April with throwing a Molotov cocktail at Sam Altman's home.

The strongest counterargument is that the displacement has not happened yet, only begun, and that the CEOs updated based on genuine observation of a labor market that has not broken the way they predicted. That is plausible. What Amodei told Axios in May 2025: "We, as the producers of this technology, have a duty and an obligation to be honest about what is coming." The S-1 tells a different story. So do the 49,000 tech workers cut this year with AI cited as the reason. And if the IPOs succeed, public shareholders will have their own questions — what exactly headcount looks like post-transition, what margins look like when the efficiency gains arrive, and whether the Jevons frame was a prediction or a promise. The quarterly earnings calendar will not wait for the rebalancing to arrive.