The AI Report That's Spooking Wall Street

A new report from MIT is casting doubt on the hype around the financial value AI brings to businesses, and may be triggering a small tech stock sell-off on Tuesday.

The report, The GenAI Divide: State of AI in Business 2025, found that the promised AI gold rush isn’t paying off for most companies yet.

Despite the major push to adopt AI tools in the corporate world, fewer than one in ten AI pilot programs have generated real revenue gains. The rest are having no impact on a company’s bottom line, according to MIT’s report—based on 150 executive interviews, a survey of 350 employees, and an analysis of 300 public AI deployments.

“Just 5% of integrated AI pilots are extracting millions in value, while the vast majority remain stuck with no measurable [profit and loss] impact,” the report said. Meaning that “95 per cent of organizations are getting zero return.”

The eye-opening report spooked investors and sent AI stocks sliding. Nvidia dropped 3.5% on Tuesday, Arm Holdings fell 3.8%, and data analytics firm Palantir took the hardest hit, plunging nearly 9% yesterday and is still sinking this morning.

Aditya Challapally, the report’s lead author, told Fortune the problem isn’t necessarily the AI models themselves, but that most companies don’t know how to use them to their full potential.

“Some large companies’ pilots and younger startups are really excelling with generative AI,” Challapally said. He highlighted startups run by 20-year-olds as good examples of how businesses should use AI. He said these startups “have seen revenues jump from zero to $20 million in a year. It’s because they pick one pain point, execute well, and partner smartly with companies who use their tools.”

Most companies, though, are misplacing their resources. More than half of generative AI budgets are funneled into sales and marketing tools, but MIT found that real returns come from boring back-office automation—things like eliminating business process outsourcing and streamlining operations.

Additionally, buying specialized tools or teaming up with outside vendors works about 67% of the time, while homegrown builds only succeed one-third as often. That’s a big deal in industries like finance, where companies are pouring money into building their own proprietary systems. MIT’s takeaway is that going solo carries more risk for failure.

The report also dropped just days after OpenAI CEO Sam Altman warned that an AI bubble may be forming.

“I do think some investors are likely to lose a lot of money, and I don’t want to minimize that—that sucks,” Altman said, according to the Financial Times. “There will be periods of irrational exuberance. But on the whole, the value for society will be huge.”

Meta shares also dipped after it announced a major shake-up in its AI division, a move some are seeing as a bad sign for the company’s AI ambitions that adds to the worries of an AI bubble.