AI Bubble? The Dream Is Starting to Crack

A few weeks ago , we left off with Mark Zuckerberg busy writing billion-dollar checks to recruit a handful of elite artificial intelligence researchers. These professionals are paid as much as—and sometimes more than—top soccer players.

The summer has passed, at least for the most part, and here we are in a completely different situation: because in the meantime, GPT-5, OpenAI's latest language model, has arrived. It should have marked a turning point, but instead it disappointed those who expected a technological leap similar to that between GPT-3 and GPT-4. "AGI," they call it. Or superintelligence. But no, it's still a chatbot, and some are starting to fear that we've reached a plateau, a phase of technological progress in which every step becomes smaller and less impressive.

Sam Altman, OpenAI's CEO, admitted it bluntly: the launch wasn't a success, especially in terms of communication . But he immediately reiterated his call, asking for "trillions of dollars"—thousands of billions of dollars—to build and upgrade the infrastructure and data centers essential to AI. And then, speaking of the danger of AI becoming a bubble, he reasoned out loud, using rhetorical questions: "Are we in a phase where investors are overly excited about AI? I think so," he said. And then: "Is AI one of the most important things that's happened in a long time? Again, I think so."

This isn't the first time Altman has made similar statements, but this time, his comments have ruffled even the most enthusiastic investors' feathers and dampened Wall Street's enthusiasm, especially regarding the hottest AI-related stocks. Even Nvidia, which recently surpassed $4 trillion in market value, has lost a few ounces; but above all, Palantir, a company that combines data analytics with national security, has lost 9% (not much, considering the company had gained over 360% in the last year, but still enough to alarm some analysts).

These data have rekindled the spirit of some voices that have long been calling for caution and reminding us that this rush to data centers and linguistic models isn't entirely sound. According to Erik Gordon, a professor at the University of Michigan, the AI ​​boom could even dwarf the dot-com bubble of the early 2000s, which has become the Valley's bogeyman.

An MIT report with the telling title, "The GenAI Divide: State of AI in Business 2025," clearly shows the gap that remains between hype and reality, especially regarding corporate adoption of these technologies. Startups and small businesses often manage to leverage AI with agility, while large companies struggle to integrate it into existing processes. Even more remarkable is the distribution of budgets: over half of spending on generative AI goes to marketing and sales, while the most solid returns come from less visible areas—logistics, accounting, and human resources management.

Despite everything, investment continues, and how: Nvidia's GPUs, data centers, and big tech stock holdings have become the engines of capital inflows that are fueling much of the US economic growth. As Axios noted, Microsoft and Nvidia alone are now worth approximately $2.5 trillion more than a year ago. These numbers are mind-boggling, considering that some of these companies (Palantir, OpenAI, Nvidia) were virtually unknown to most until a few years ago and today have valuations higher than the GDP of some G8 countries.

All of this, however, rests on a promise: that artificial intelligence will change the world quickly, so quickly that it will justify the construction of hundreds of data centers, the massive use of energy and water to cool them, and investments on a previously unheard of scale. It's a fascinating promise, at least for some, and undoubtedly profitable, at least so far; but it's also increasingly risky.

Because the gap between reality and the incredible near future painted by Altman & Co. is widening, while AI innovations continue to arrive in a constant cycle, perhaps even numbing us. The surprise effect has faded, leaving only confusion and concern about the job market. And if we truly are facing a bubble, the danger would be compounded by the fact that, this time, unlike the dot-com era, AI appears to be driving (or drugging) the entire US economy.

More on these topics: