Bezos promises a labor shortage. The numbers say otherwise.

Bezos promises a labor shortage. The numbers say otherwise.

On the VivaTech stage on June 17, Jeff Bezos waved away the most widely shared fear of the moment. A lot of smart people worry that AI will make humans superfluous, he acknowledged, before flatly disagreeing: in his view, AI will instead create "a labor shortage." Not unemployment. A shortage. The exact opposite.

At that very moment, the tech layoff counter showed 142,000 jobs cut since January 2026. The gap between the sentence and the figure is the real story.

What Bezos says, and why it holds up

The argument isn't absurd. For Bezos, human wants are endless. What holds us back are barriers: cost, time, complexity. AI tears down those barriers, so we take on more, so we need more hands.

He illustrated it in May on CNBC: if you've been digging out your home's basement with a shovel and someone hands you a bulldozer, you should be thrilled. The tool doesn't replace the worker, it multiplies what he can do, and you hand him jobs no one would ever have started with a shovel. Bezos brushes aside the fears of radiologists and developers along the way.

It's a classic reading of technical progress. The power loom didn't kill textiles, it made them explode. The argument has two centuries of history behind it, and it has often been right.

Except the 2026 data tells a different story

The problem is the present. The 142,000 jobs cut in tech through the end of May don't come from struggling companies. Meta, Amazon and Oracle are laying off while posting record results and pouring 700 billion dollars into AI infrastructure. Oracle cut up to 30,000 jobs on April 1. Meta notified 8,000 departures on May 20, Intuit 3,000 the same day.

Goldman Sachs estimates that payroll cuts directly attributed to AI run to more than 16,000 a month in 2026 among large U.S. employers. The bank sums up its read in three words: "automation over hiring."

And it's accelerating. A survey by Duke University and the Atlanta and Richmond Federal Reserve banks, of 750 chief financial officers, anticipates AI-related layoffs nine times higher in 2026 than in 2025.

Economists' projection stays measured, around 0.4% of U.S. employment, or roughly 500,000 jobs. John Graham, who runs the survey, points out himself that "this is not the apocalyptic scenario of the headlines." But 500,000 is still the opposite of a labor shortage.

One caveat applies to these figures: the attribution to AI is often self-reported. A company laying off to cut costs has every reason to dress the decision up as "AI transformation" rather than admitting it was overstaffed. These numbers probably overstate technology's real role. They still point overwhelmingly downward.

Where the man promising jobs is speaking from

Bezos didn't say this as a neutral bystander. He was at VivaTech to talk about Prometheus, the AI startup he co-founded in November 2025 with Vik Bajaj, a Google X alum. The round: 12 billion dollars raised, for a valuation of around 41 billion. The product: what Bezos calls an "artificial general engineer," a design tool for industry, aerospace, automotive and pharma.

In other words, an AI whose explicit purpose is to do the work of engineers and technicians faster, with fewer people. Promising a labor shortage while selling the tool that automates engineering is a position, not a disinterested forecast. It doesn't prove he's wrong. It explains why his optimism lands exactly where it's convenient.

The real question isn't who's lying, it's when

The trap would be to conclude that Bezos is talking nonsense. The truth lies in a distinction he glosses over: the time scale.

Short term, the numbers prove him wrong. Automation is cutting jobs now, faster than it creates new ones, and the people losing their junior developer roles don't become "bulldozer-assisted engineers" overnight. A Reuters/Ipsos poll in early June confirms it on the sentiment side: 53% of Americans worry about their career prospects because of AI.

Long term, his thesis stays open. Economists know the Jevons paradox: when a technology makes a resource cheaper, we consume more of it instead of saving. Applied to labor, that gives Bezos's intuition. If designing an object costs ten times less, we design ten times more objects, and we may need more people overall. Add the aging demographics of wealthy countries, which promise genuine shortages of hands, and the idea of a shortage twenty years out is hardly far-fetched.

So Bezos's claim isn't false. It's just offset by a decade or two from the present he's commenting on. And that's convenient: talking about the industrial revolution in the past tense is a way to avoid looking at this quarter's layoff sheets.

The same blind spot, French edition

This comfort isn't a Silicon Valley specialty. The same day Bezos was touting his shortage, word came that Matignon was sitting on a cross-inspectorate report on AI's impact on public-sector headcount. The decision, taken at the highest level: don't publish it. The inspectors themselves were asking for its release, and the figures obtained reportedly weren't even alarming.

A billionaire promising abundance to sell his AI, a government burying reassuring data out of political caution: both refuse to put the numbers on the table, for opposite reasons. That may be the real signal of 2026. Everyone has a take on AI and jobs. Almost no one wants to show their books.