Every gold rush has its true believers and its skeptics.
The believers see a new frontier, infinite upside, a future where the old rules no longer apply. The skeptics see the same hills, the same picks, and a lot of people pretending the math works out.
Silicon Valley has spent three years acting like the believers won that argument. Companies have rebranded as AI-first, retooled their org charts, and started measuring productivity in agents rather than people. Boards have rewarded the rhetoric with valuations that would have looked absurd in 2023.
The layoffs followed. So did the corporate press releases explaining that the cuts were not really cuts at all, just a rational response to a technology that had finally arrived.
Boards approved them. Analysts upgraded the stocks. Workers updated their resumes.
Now one Silicon Valley CEO is publicly calling the whole posture a clinical problem.
Photo by Bloomberg on Getty Images
Box CEO Aaron Levie diagnoses ‘AI psychosis’ in his fellow tech executives
His name is Aaron Levie. He founded Box (BOX), the cloud content management company, in 2005 and has spent the last two decades arguing that software is going to eat almost everything. He is, by any reasonable definition, an AI bull.
That is what made his post on May 24 land the way it did.
More AI:
- Micron sits at the center of a red-hot chip rally
- IBM CEO sends blunt message on AI and quantum computing
- Anthropic CEO makes shocking admission about AI
“CEOs are uniquely prone to AI psychosis because they’re sufficiently distant from the last mile of work that still has to happen to generate most value with AI,” Levie wrote on X (formerly Twitter).
His argument is uncomfortable in its simplicity. CEOs play with AI tools. They generate a contract draft. They watch a demo. They make the leap to believing agents can quietly take over the underlying jobs.
But CEOs are not the people who have to review code, find calls to hallucinated libraries, or comb through hundreds of pages of contract terms to flag what the model missed.
Levie’s diagnosis is not anti-AI. He is an active angel investor in AI startups and posts AI optimism to 2.7 million followers on X. His point is narrower and more uncomfortable. The people approving the layoffs do not know what the underlying work actually is.
Related: Uber CEO, COO sends stark message on AI spending in 2026
Tech industry layoffs hit a grim 2026 milestone in only five months
The receipts are easy to find. In the first five months of 2026 alone, 115,430 people have been let go from 152 tech companies, industry tracker Layoffs.fyi reported. That figure is already closing in on the entire 2025 total of 124,636 people from 275 companies.
When I ran my own analysis against the 2025 baseline, the pace is brutal. The industry is on track to fire more people this year than in any year since the dot-com bust, and almost every press release blames the same culprit. The culprit is AI productivity.
Most of the time, the math does not check out. Many large tech companies are “AI washing,” meaning they are crediting AI productivity for cuts actually driven by overhiring, slowing revenue, or basic cost discipline, TechCrunch reported in February.
A clean example landed on May 25. ClickUp founder Zeb Evans publicly announced he had laid off 22% of his staff after rolling out roughly 3,000 internal AI agents, TechCrunch reported. Evans framed the cuts as the start of a “100x org,” not a cost move.
That framing has become standard. The numbers underneath it have not.
The research case against the AI productivity story keeps growing
Here is where the story gets uncomfortable for the executives doing the cutting.
What four major studies actually found
- A meta-analysis published in October 2025 found “no robust relationship between AI adoption and aggregate productivity gain,” according to UC Berkeley’s California Management Review.
- A March 2026 working paper concluded AI adoption did improve productivity, but documented a “productivity paradox” in which “perceived productivity gains are larger than measured productivity gains,” the National Bureau of Economic Research reported.
- AI agents are not yet producing human-quality work on most tasks and may not consistently do so until 2029 or later, MIT FutureTech researchers concluded.
- When everyone uses AI to produce more output, the bottleneck simply moves to the executives who must approve all the new output, the Harvard Business Review reported in May.
Read those four findings in sequence and a pattern emerges. The productivity gains every CEO is racing to capture are small, mismeasured, or several years away from arriving at the scale required to justify the layoff totals already on the board.
What the Silicon Valley AI delusion means for your portfolio and your job
So what should you do with all this if you own tech stocks, work in tech, or are saving for retirement through an index fund that is now roughly 35% Magnificent Seven by weight?
What struck me as I read through the research is how lopsided the trade has become. The market is pricing in productivity gains that the researchers studying productivity cannot actually find. Gaps that wide tend to close, and they tend to close painfully.
The smart money is positioning around exactly that risk. Investor Michael Burry has built a public anti-AI bet on this exact thesis. Peter Thiel trimmed his top AI position, stirring fresh bubble talk. Stanley Druckenmiller has quietly rotated into less obvious tech names.
For workers, the implication is more direct. The “forever layoffs” era is not winding down. It is restructuring around a story that may not hold past 2029, the earliest date MIT’s researchers believe agents will reliably do most text work at human quality.
Levie left his peers a way out. Use AI heavily, he advised, and “come out the other side with an appreciation for both the upside and the real work,” he wrote on X.
The CEOs who actually follow that advice will be the ones still standing when the agents finally start doing the job they have already been paid to do.
Related: Pope Leo XIV teams with Anthropic founder on bold AI warning