The Google DeepMind CEO sees a 'new golden era of discovery' ahead but flags bubble dynamics in AI startup valuations
Demis Hassabis, co-founder and CEO of Google DeepMind, described artificial intelligence as “overhyped in the short term” but “underappreciated in the medium to long term.” He predicted a “new golden era of discovery” across medicine, energy, and science over the next 10 to 15 years.
The bubble warning from inside the machine #
Hassabis flagged the risk of an “over-correction” in AI startup valuations. Massive amounts of capital are flowing into pre-revenue AI companies, and the investment landscape swung rapidly from skepticism to obsession.
Hassabis has been voicing similar concerns since 2024, consistently emphasizing that achieving true artificial general intelligence, or AGI, requires additional breakthroughs beyond simply scaling up existing models. He places the AGI timeline at approximately 2030, give or take a year, based on remarks he reiterated during talks at Stanford in June 2026.
Why the long-term optimism matters more than the short-term warning #
DeepMind’s AlphaFold project essentially solved the protein-folding problem, a challenge that had stumped biologists for 50 years. The sectors Hassabis highlighted—medicine, energy, and broader scientific research—represent trillions of dollars in global economic activity.
What this means for investors navigating the AI landscape #
Pre-revenue AI startups commanding enormous valuations are vulnerable to exactly the kind of over-correction Hassabis described. For the crypto market specifically, Hassabis made no mention of tokens, digital assets, or decentralized protocols in his remarks. If AGI or something approaching it arrives around 2030, the infrastructure and applications built on top of it will take additional years to mature, suggesting the most significant investment opportunities in AI may not peak until the early-to-mid 2030s.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our