An MIT Report Finds 95 Percent of Company AI Projects Show No Profit

A study from an MIT lab found that the vast majority of business efforts to deploy generative AI had produced no measurable return, even as spending soared. Coming amid record valuations, it sharpened the question of whether AI is a bubble.

A report from an MIT research initiative, widely circulated this week, found that about 95 percent of corporate projects using generative AI had produced no measurable financial return. The study looked at hundreds of company deployments and concluded that the great majority were stuck — pilots that never turned into anything, tools that employees didn't adopt, projects that cost money without saving or making it. The successes clustered in narrow uses, often unglamorous ones like automating a specific back-office task, rather than the sweeping transformations companies had announced.

The finding is not that the technology doesn't work — individuals use these tools productively every day — but that turning it into business value at scale is proving hard, slow, and often disappointing, the way most new office technology has been. The gap between what a chatbot can do in a demo and what it reliably does inside a large organization's messy systems turned out to be wide.

The timing gave it weight. It arrived as AI company valuations hit records, as Nvidia crossed $4 trillion, and as hundreds of billions were being committed to data centers on the premise that AI would soon transform the economy. A report saying the transformation was, so far, mostly not happening was exactly the counterweight the skeptics had been waiting for, and it moved markets for a few days.

Both things can be true: that the technology is genuinely capable, and that most attempts to profit from it so far have failed. Which one the future belongs to — the capability or the disappointment — is the question the word "bubble" is really asking, and this report put a hard number on the disappointing half.