The sweeping layoffs in Silicon Valley are proof that tech executives are suffering from ‘AI psychosis,’ says Box’s CEO.
There is a growing disconnect in Silicon Valley between the corner office and cubicles.
In a recent letter to X, Aaron Levie, CEO of the content management platform, said the silent part loudly about how his peers in the tech world are failing to grasp the full extent of AI’s work.
“CEOs are prone to AI psychosis because they are far from the last mile of work that needs to be done to create great value with AI,” Levie wrote in X.
He added: “So when they’re playing with AI, they’re seeing results in a fun way, often not thinking about the next 10 or 20 things that have to happen to get sustainable results for agents.”
In other words, CEOs see only the best in technology, away from the bugs, bad ideas, and other snafus employees are tasked with encountering every day.
That view shows what is seen in the data. A 2025 study from AI firm Rev found that heavy AI users encountered three times as many ideas and spent nearly 10 times as long getting answers. Those who work are “tokenmaxxing,” or maximizing the amount of AI tokens they burn. That’s part of the technology that some CEOs simply fail to understand as they plan to lay off thousands of workers to install AI.
And ROI projections are even worse. A recent Gartner survey of 350 global executives with annual revenue of at least $1 billion found that while 80% of those who had tested AI or autonomous technology reported downsizing, businesses cut jobs regardless of whether the technology generated returns. In other words, AI-related layoffs may be for show rather than impact. Many business leaders such as OpenAI CEO Sam Altman argue that companies are washing AI, or using the technology as a valid reason for layoffs when in fact, they are done because of another basic disease of the budget.
The token tab goes up
Levie gives an example of how this AI problem plays out at work.
“‘Look I’ve made a contract,'” an employee might announce. He continues: “But you didn’t verify all the terms before it went out to the other party and you didn’t have to connect all the previous contracts to work with them. There could be a gap between the envisioned end product and the work required to achieve that.”
That’s becoming an expensive consideration because the AI tabs are starting to deteriorate. Microsoft recently switched from Claude to GitHub Copilot CLI. Uber CTO Praveen Neppalli Naga recently told Information the company had spent its annual budget for 2026 AI coding tools in just four months. And an AI consultant client Axios recently spoke to said it misappropriated half a billion dollars after failing to impose usage restrictions on Claude’s licenses to employees.
