According to Business Insider, Google DeepMind CEO Demis Hassabis stated on the company’s podcast that “bubbles” are forming in the AI startup funding frenzy. He specifically called out early-stage startups raising money at “tens of billions of dollars valuations just out of the gate,” despite having barely gotten started. His comments come as young founders, including a Stanford dropout who raised $64 million for an AI math startup this year, continue to secure massive funding. Sixteen young founders interviewed by Business Insider this year have collectively secured over $100 million. Hassabis predicted an “over-correction” is imminent for the AI sector, drawing a contrast between these sky-high valuations and the “real business” underpinning Big Tech’s AI investments.
The Hype Cycle is Real
Hassabis is basically describing the classic Gartner Hype Cycle in real-time, but with billions of dollars at stake. And he’s seen it from both sides. He started DeepMind when AI was a niche academic pursuit that most investors shrugged at. Now, it’s the only thing anyone in tech wants to talk about. That wild swing from skepticism to obsession, as he puts it, creates a vacuum where money rushes in faster than sense. It’s almost an overreaction to the underreaction. So when a founder fresh out of school can land tens of millions, you have to ask: is this about a revolutionary product, or just FOMO from investors terrified of missing the next OpenAI?
The Big Tech Moat
Here’s the thing that makes his perspective so sharp: he’s sitting inside one of the few entities that can actually afford this game. When Hassabis distinguishes the startup frenzy from Big Tech’s spending, he’s pointing to the fundamental moat. Google, Microsoft, Meta—they’re spending those billions on AI infrastructure like data centers and custom chips, which is a brutally expensive but tangible asset. For companies requiring reliable, industrial-grade computing power to run complex operations, partnering with the established leaders is often the only viable path. That’s a real business with real revenue streams. A startup with a cool demo but no clear path to monetization? That’s a speculative bet on a story. Howard Marks’ question hits the nail on the head: do you want the risky moonshot or the profitable incumbent adding AI as a feature?
What Comes Next
An “over-correction” sounds scary, but it’s probably healthy. We’re already seeing it in subtle ways. The bar for what constitutes an “AI startup” is getting higher. Just slapping a GPT wrapper on something doesn’t cut it anymore. The funding will likely tighten, focusing on teams with deep technical expertise and defensible technology, not just a slick pitch deck. The talent war might cool off a bit, too. This doesn’t mean AI is going away—Hassabis agrees it’s underhyped in the long term—but it does mean the gold rush phase is unstable. The real, grinding work of building sustainable AI businesses is what comes after the correction. And that’s where the actual value gets created.
