The numbers are staggering. Tech giants are pouring tens of billions into artificial intelligence, while AI startups command valuations that would have seemed fantastical just years ago. Price-earnings ratios have reached levels reminiscent of only one other period in modern history: the late 1990s tech bubble.
The parallels are hard to ignore. Like the Information and Communication Technology boom of the late ’90s, today’s AI revolution combines genuine technological breakthroughs with abundant capital and fierce competition for talent. And like then, the prevailing wisdom suggests that even if we’re in a bubble, the long-term benefits will outweigh the short-term excesses. After all, the ICT boom left us with faster internet, e-commerce, and the foundation for today’s digital economy.
But new research from economists Johan Hombert and Adrien Matray challenges this comforting narrative. Their study of the ICT boom reveals a troubling pattern that should give pause to anyone celebrating the current AI investment frenzy: the workers who joined the booming sector often ended up worse off in the long run.
Tracking thousands of high-skilled French workers who entered the job market during the late 1990s, the researchers found that those who joined ICT companies during the boom earned significantly less 15 years later than comparable workers who started in other sectors—despite initially commanding premium wages. The wage discount was substantial: about 7%, equivalent to losing two years of career progression.
This wasn’t about picking the wrong companies. Even workers who joined successful firms that weathered the post-boom downturn faced the same wage penalty. Nor was it simply a matter of the sector experiencing hard times—workers who joined ICT companies just a few years after the boom ended showed no such decline.
The culprit appears to be rapid skill obsolescence. The intense experimentation and fast-paced innovation that characterize technology booms create a double-edged sword. Workers gain experience with cutting-edge technologies, but those specific skills quickly become outdated as newer approaches emerge. Software developers who specialized in building static websites with early HTML in the late 1990s found their expertise less valuable once database-driven technologies became standard. IT consultants who implemented on-premise systems saw their knowledge depreciate as cloud-based solutions took over.
“Skills accumulated in boom-era ICT jobs became obsolete unusually quickly, as technologies introduced during the boom were rapidly replaced,” says Adrien Matray. “Workers specialized in these transient technologies saw the value of their human capital erode.”
The research reveals that this effect was concentrated among STEM workers—engineers, developers, and technical specialists whose skills were tightly coupled to specific technological implementations. Meanwhile, workers in more general roles like finance or management showed no such wage decline, even at the same firms.
Perhaps most troubling, the study found that capital flowed disproportionately toward firms where workers’ skills would depreciate fastest. Companies engaged in the most intensive experimentation—those most likely to generate multiple generations of quickly obsolescent technology—attracted the largest investments. This means the financing boom didn’t just expose more workers to skill obsolescence; it amplified the effect by directing them toward the riskiest positions.
The implications for today’s AI boom are sobering. The sector is experiencing even more intense experimentation and faster technological change than ICT did in the late 1990s. AI frameworks, tools, and best practices evolve by the month. Large language models released as state-of-the-art become outdated within a year. Skills in one generation of AI technology may prove worthless for the next.
The researchers estimate that one-third of skilled workers entering the French labor market during the ICT boom joined that sector. Today’s proportions in AI-related fields may be smaller, but the absolute numbers—and the capital deployed—are far larger. If history repeats itself, we may be training a generation of highly skilled workers for careers built on sand.
This doesn’t mean the AI boom is bad for society overall. The technologies being developed may indeed transform productivity and create enormous value. But we shouldn’t assume that channeling talent and capital into a booming sector automatically benefits those workers in the long run. The invisible hand of the market, when combined with speculative financing and rapid technological change, may point talented individuals toward careers that offer glittering entry packages but dim long-term prospects.
For policymakers, this research suggests that labor market impacts of investment booms deserve more attention. For workers contemplating careers in hot sectors, it’s a reminder that premium entry wages may reflect risk as much as opportunity. And for all of us watching the AI revolution unfold, it’s a cautionary tale about assuming that more investment and more innovation automatically translate into broadly shared prosperity.
The dot-com boom taught us that bubbles can burst. This research shows that even before they do, they may already be eroding the human capital of those caught up in the excitement.