Demis Hassabis, the cofounder and current chief executive officer of DeepMind, has delivered an unvarnished message directed at certain corners of the rapidly expanding artificial intelligence startup ecosystem: in his estimation, much of what he sees unfolding appears precarious and unlikely to endure over time. During a recent episode of *Google DeepMind: The Podcast*, released on Tuesday, Hassabis reflected on the current state of AI investment and pointed to unmistakable signs that speculative excess—what he termed funding “bubbles”—may be forming amid the frenzy surrounding the technology. He specifically noted that this phenomenon seems most evident among early-stage startups that are attracting extraordinarily high valuations, often amounting to billions of dollars, despite having little more than initial concepts or untested prototypes to show for their efforts.
According to Hassabis, some companies are managing to secure financing at valuations in the tens of billions, even though, in his words, they “basically haven’t even got going yet.” The mere existence of such astronomical valuations at such an embryonic stage, he suggested, raises important questions about long-term viability and the underlying realities of the marketplace. With characteristic candor, Hassabis questioned how such a pattern of exponential financial optimism could possibly be sustained, ultimately concluding that, in most cases, it probably cannot. The implication of his observation is not an outright dismissal of innovation but rather a sober assessment that enthusiasm, left unchecked by fundamentals, often leads to inevitable correction.
He was careful, however, to draw an important distinction between these fledgling startups and the well-established technology giants currently channeling billions into developing and maintaining massive AI infrastructure. Unlike the speculative valuations of startups that have yet to deliver tangible products or consistent revenue streams, the valuations of Big Tech firms, Hassabis explained, rest upon “a lot of real business.” In other words, the leading corporate players in AI—notably companies such as Google, Microsoft, and others—possess profitable operations, mature technological foundations, and substantial user bases that lend genuine credibility and durability to their financial figures.
Reflecting on the current moment, Hassabis characterized AI as an innovation cycle that tends to be “overhyped in the short term” yet simultaneously “underappreciated in the medium to long-term.” What he meant by this, he elaborated, is that new technologies often experience a burst of inflated expectations when they first capture the public imagination, only to be underestimated later when their true transformative potential becomes apparent. The history of technological revolutions—from the internet to mobile computing—has followed this same rhythm of initial excitement, overcorrection, and eventual stabilization.
Hassabis also predicted that this pattern of volatility would repeat itself with AI. He used the term “over-correction” to describe what typically happens as an industry matures from skepticism to widespread obsession in a matter of years. To illustrate, he recalled the early days of DeepMind, observing that when the company first launched, virtually no one believed in the feasibility of what his team was attempting to accomplish. Fast forward a decade and a half, and artificial intelligence has transitioned from being a niche research pursuit to a central topic in every major business conversation. That pendulum swing—from disbelief to all-consuming fascination—often drives valuations upward too quickly, producing what Hassabis referred to as “an overreaction to the underreaction.”
Despite widespread public debate about whether the current AI boom represents a full-fledged bubble, Hassabis clarified that he does not spend time worrying about such labels. His focus, he emphasized, remains on leading DeepMind’s research efforts and ensuring that the company continues to build advanced AI systems that contribute directly to Google’s products and services. DeepMind’s work underpins many of Google’s most complex models, including Gemini, and remains at the forefront of fundamental AI research and safety exploration.
The remarks from Hassabis arrive at a time when AI startups around the world are attracting staggering multiples of investment. In numerous cases, founders barely out of their university programs are raising tens of millions of dollars, drawing not only capital but also elite researchers and engineers from leading institutions like Meta and Google Brain. For example, Business Insider recently reported that a Stanford dropout managed to raise $64 million for her fledgling AI mathematics company, Axiom Math, successfully persuading top-tier investors and recruiting seasoned AI scientists from established tech laboratories. In total, sixteen young founders profiled this year collectively amassed more than $100 million in funding—a figure that underscores investor appetite but also highlights the speculative intensity of the market.
Nevertheless, not everyone within the investment community is convinced that this extraordinary surge in funding is justified. Howard Marks, billionaire cofounder of Oaktree Capital Management, echoed similar caution during a separate interview on the *We Study Billionaires* podcast. Marks observed that investor enthusiasm seems to have outpaced the evidence, with money flowing into AI startups that display minimal operational history and offer little proof of sustainable profitability. He posed a sober rhetorical question to investors: Would they prefer to gamble on a new, untested startup that currently generates no revenue and no earnings but carries the potential—however remote—for astronomical success if its technology proves revolutionary? Or would they rather place their trust in an established technology titan already generating substantial profit, where artificial intelligence might add value incrementally but not transform the entire business model?
This, Marks suggested, is the critical choice confronting investors today: between unbounded speculative possibility and dependable, incremental innovation. Taken together, the voices of both Hassabis and Marks point to a broader reality emerging within the AI industry—a moment characterized by unprecedented creativity and ambition but also by unmistakable signals of exuberance that may soon be tempered by the practical forces of market correction.
Sourse: https://www.businessinsider.com/demis-hassabis-google-deepmind-ai-startup-valuation-correction-bubble-2025-12
