According to New Atlas, Google DeepMind co-founder Demis Hassabis admitted in a Hard Fork podcast interview that parts of the AI industry are “probably in a bubble,” specifically pointing to “multi-ten-billion-dollar” seed rounds for companies with “basically nothing.” His comments came alongside Alphabet CEO Sundar Pichai’s warning that “no company is going to be immune” from market corrections, with both executives speaking as Google launched its Gemini 3 model. The bubble concerns come as OpenAI reportedly lost $14 billion this quarter alone, while Nvidia became the first company to surpass $5 trillion valuation before facing major short sellers like Michael Burry. Hassabis remains optimistic about Alphabet’s position, citing potential in “half a dozen to a dozen” new business areas like drug discovery and robotics that could become “massive multi-hundred-billion-dollar businesses.”
The bubble warnings are getting louder
Here’s the thing – when the people building this stuff start saying “bubble,” you should probably listen. Hassabis isn’t some outside critic; he’s literally running Google‘s AI division. His specific concern about “multi-ten-billion-dollar rounds with basically nothing” hits hard because we’ve all seen those funding announcements. It’s basically the dot-com era all over again, where companies were valued on potential rather than actual revenue.
And the numbers are staggering. OpenAI burning through $14 billion in a single quarter? That’s insane when you think about it. Meanwhile, Nvidia hitting $5 trillion valuation only to immediately face short sellers like Michael Burry – the guy who famously predicted the 2008 financial crisis – suggests the smart money is starting to get nervous. These aren’t small players getting cold feet; Burry made $100 million betting against subprime mortgages, so when he starts shorting AI stocks, people notice.
Who actually survives a bubble burst?
Hassabis makes a compelling point about Alphabet’s position. Basically, they’re trying to win whether there’s a bubble or not. The immediate returns come from integrating AI into existing “multi-billion-user products,” while simultaneously investing in future moonshots. It’s a hedge that smaller pure-play AI companies can’t really match.
Look at the diversity he mentions – drug discovery through Isomorphic Labs, robotics, gaming, NotebookLM. That’s the advantage of being a tech giant versus a startup betting everything on one AI application. When the inevitable shakeout comes, companies with multiple revenue streams and established user bases will weather the storm much better than those depending entirely on speculative AI investments.
We’ve been here before
The dot-com comparison isn’t just lazy journalism – Pichai himself used the term “irrational exuberance” that defined that era. According to business cycle data, the 2001 recession was brief but brutal for tech employment. The difference this time? AI actually has transformative potential across multiple industries, from scientific research (Hassabis literally won a Nobel Prize in Chemistry for AlphaFold) to manufacturing automation.
But potential doesn’t pay the bills today. The question is whether current valuations reflect realistic near-term revenue versus speculative future possibilities. When seed rounds hit billions for pre-revenue companies, that’s usually a red flag. The scary part? Former IMF chief economist Gita Gopinath thinks an AI crash could wipe out $20 trillion from US households. That would make the dot-com bust look mild.
So what happens now?
Hassabis is probably right that a bubble burst would “thin out the crowded AI industry.” We’d likely see consolidation, with giants like Alphabet, Microsoft, and Amazon scooping up promising technology from failed startups at fire-sale prices. The employment impact would be immediate and painful for the tech sector.
But here’s the silver lining – the internet survived the dot-com crash, and the companies that emerged were stronger and more focused. AI isn’t going away, but the business models might need to mature. As for Gemini 3 and other near-term products, they’ll need to demonstrate real utility rather than just hype. The next year will show whether this is truly transformative technology or just the latest speculative frenzy.
