Opinion: Don’t Fear the A.I. Bubble Bursting
By Carl Benedikt Frey | December 5, 2025
Published in The New York Times
As the hype surrounding artificial intelligence (A.I.) reaches fever pitch, many fear that the sector is headed for a damaging bubble burst. But according to Carl Benedikt Frey, an economist at the University of Oxford and author of How Progress Ends: Technology, Innovation and the Fate of Nations, such a deflation may actually be beneficial for the future of A.I. development.
The Role of Bubbles in Innovation
Eric Schmidt, former CEO of Google, has said, “Bubbles are great. May the bubbles continue.” This reflects a widespread belief in the tech industry that enormous investment and boundless resources are essential to push A.I. breakthroughs — from curing cancer to achieving true artificial general intelligence (AGI) and surpassing global competitors like China.
The conventional story suggests that the way forward is to build more data centers, buy faster chips, and pour billions into ever-expanding infrastructure. But Frey challenges this assumption, arguing that breakthroughs often occur under pressure rather than in times of abundance. When resources become scarce or expensive, innovation is directed toward efficiency and smarter solutions.
Innovation Through Hardship: Lessons from History
Frey invokes historical examples to demonstrate how crises can spur critical technological advances. During the 1970s energy crisis, when Americans lined up for gasoline, businesses and governments responded by improving energy efficiency across industry, housing, and transportation. The result was a wave of innovations still impacting us today, including more efficient engines, better insulation, and the emergence of electric and hybrid vehicles.
Earlier still, the catastrophic 1927 Mississippi River flood devastated agricultural communities, severely reducing available labor. With fewer hands to plant and harvest crops, farmers rapidly mechanized, embracing tractors and mechanical tools at a pace unmatched in neighboring areas. This labor scarcity directly drove agricultural modernization.
The Need for a New Approach in A.I.
Frey contends that generative A.I. currently suffers from similar challenges. Large language models, impressive as they are, are fundamentally limited. These models primarily predict what a human might say next based on the data they’ve been trained on; they do not “think” or innovate independently. Training a model on outdated texts from the 1800s, for example, won’t yield innovations like airplanes or rockets—it simply regurgitates the ideas of that time.
Merely scaling up hardware with faster chips and bigger data centers isn’t enough. Such an approach is costly, energy-intensive, and risks plateauing on today’s “limited, mediocre” results.
Energy Efficiency and Machine Learning: The Critical Frontier
For A.I. to truly advance, it must evolve to learn continuously and efficiently. Unlike current models that need to be retrained from scratch and do not retain learning during operation, the next generation of A.I. should emulate human memory and learning habits, improving and innovating on the fly.
Frey envisions technology that harnesses greater computational power with less energy expenditure, allowing models to do more work per watt consumed. Such advancements would mark a leap closer to replicating human cognitive flexibility and creativity.
Conclusion: The Silver Lining in a Cooling Market
Rather than fearing the contraction of funding and the cooling of the A.I. investment boom, Frey suggests embracing it as a necessary “course correction.” Scarcity and constraint may be the very catalysts driving smarter, more sustainable innovation in the field. Past history shows that the greatest technological progress often comes when we have no choice but to do better with less.
As the current A.I. bubble deflates, the industry faces a crucial opportunity to rethink its priorities—pivoting from mindless expansion to purposeful efficiency and genuine technological breakthrough.
Carl Benedikt Frey is an economist at the University of Oxford and author of the book “How Progress Ends: Technology, Innovation and the Fate of Nations.”
