Why the MIT AI Report Misreads the Market: Early Failures Don’t Mean

Introduction: Publish or Perish

In academia, there is a common refrain: “Publish or perish.” I am well aware of this pressure from my own academic background, including my time at Stanford University, and from my role as Editor-in-Chief of the scholarly Journal of Active and Passive Electronic Devices (Old City Publishing), where academics from around the world submit their articles for editing before publication. The imperative to publish often means that research is presented in ways designed to stand out, even when its conclusions are more about attracting attention than providing balanced insight.

The recent MIT report claiming that 95 percent of companies see zero returns from their generative AI investments is a prime example of this. The study made headlines, triggered a sudden market sell-off, and raised questions about whether AI is already in bubble territory. But as history shows us, such reports tend to measure transformative technologies against the wrong yardstick: short-term profitability.

Technology Diffusion and Early Low Returns

History demonstrates that every major technological revolution began with an extended period of disappointment before it transformed industries. When electricity was first introduced into factories, productivity gains were negligible because plant layouts remained tied to the architecture of steam power. Only when managers redesigned production lines did electricity’s potential become visible — a process that took nearly two decades. Computers suffered the same fate: installed in offices during the 1960s and 1970s, they were initially glorified calculators that did little to change how work was done. It took a full 15 years before their integration into enterprise workflows began to pay measurable dividends.

The internet, too, endured its skepticism. The 1990s produced thousands of websites with no business models, culminating in the dot-com crash. Yet within a decade, online commerce, advertising, and communications had become indispensable. Cloud computing went through a similar phase in the 2000s — dismissed as an expensive experiment with no ROI until scale, standardization, and enterprise trust unlocked its true value.

According to Table 1, generative AI fits squarely into this historical sequence. The MIT report interprets today’s lack of immediate ROI as evidence of futility, but in truth, it is simply evidence that AI is at the same stage electricity, the internet, and the cloud once occupied: early, awkward, and misunderstood, yet on the cusp of exponential impact. Just as Amazon and Google grew from internet “failures,” and AWS grew from the cloud’s early skepticism, companies like Nvidia, Microsoft, Meta, and Google are laying the groundwork for AI’s inevitable maturation.