U.S. technology stocks with strong ties to artificial intelligence slid on Tuesday after a new report raised questions about the commercial impact of generative AI projects. The NASDAQ Composite closed down 1.4%, its sharpest single-day drop since early August. Shares of Palantir fell 9.4%, while Arm Holdings dropped 5%.
The decline followed the release of a study from NANDA, an AI research group spun out of MIT’s Media Lab, which found that only 5% of generative AI pilots move into production and generate measurable financial value. The report analyzed more than 300 public AI initiatives, 52 structured interviews with enterprise decision-makers, and a survey of 153 executives. It tracked project performance for six months after pilot stage.
According to the findings, most generative AI initiatives fail to impact company profit and loss metrics. While organizations often deploy AI in customer-facing areas, the highest success rates were observed in back-office operations, where automation reduces reliance on third-party agencies and business process outsourcing. However, the study found little effect on overall staffing levels.
Despite limited enterprise-level success, employees reported personal productivity benefits. Ninety percent of staff said they had gained from using publicly available AI tools like large language models, though only about 40% of companies pay for subscriptions.
A recurring challenge for failed projects was AI’s lack of contextual awareness. Interviewees highlighted frustrations such as “[The AI system] doesn’t learn from our feedback” and “Too much manual context required each time.” NANDA suggested that organizations partnering with vendors who can deliver adaptive, learning-capable AI are more likely to succeed.
By sector, media and telecom reported the most positive outcomes from generative AI deployments, followed by professional services, healthcare, consumer retail, and financial services. Adoption in energy and materials remained negligible. Within organizations, sales and marketing dominated AI use cases, while finance and procurement lagged far behind.
The report also noted a mismatch between tasks assigned to AI and human expectations. Managers entrusted AI with straightforward functions such as summarizing reports or drafting emails but rarely for complex responsibilities like client management.
Critics noted that the report’s marketing-heavy tone raised questions about its objectivity, particularly as NANDA positions itself as a potential vendor in the same space. Still, its headline finding—that the vast majority of generative AI projects fail to deliver tangible financial returns—appeared to resonate with investors, triggering caution in the markets.
