Demis Hassabis, CEO of Google DeepMind and 2024 Nobel laureate in Chemistry, has spent the first months of 2026 articulating a specific and measurable thesis: AlphaFold, the protein-structure model that now serves more than three million researchers across 190 countries, is the first rung in an AI-powered scientific revolution, not the last. His clearest statement of this position came in a Semafor interview published 21 January 2026, where he described the commercial boom in generative AI as potentially slowing, not accelerating, progress toward deeper scientific breakthroughs.
The Paradox at the Heart of AI Progress
In his January 2026 Semafor interview, Hassabis offered an assessment that sits at odds with the prevailing industry narrative. "The paradox of AI progress," he said, is that the commercial success of generative AI may actually stretch out the timeline to whatever comes after it. Shortages in high-bandwidth memory and reduced open research sharing have created friction in scaling, which he acknowledged as potentially useful: "It may be a good thing that it's not as fast. There's a whole bunch of other things that we need to think through with this technology." He noted that commercial pressure has made it harder to share research openly, describing it as "a shame on the one hand, but understandable."
That framing is deliberate. Where competitors have staked out exponential-scaling positions, Hassabis is positioning DeepMind's approach around multimodal training data and reasoning rather than raw compute volume. "Text alone would not get you to the endgame faster," he told Semafor, explaining DeepMind's bet that reasoning would emerge through richer, multi-modal inputs. The implication is that DeepMind's path diverges from the brute-force scaling playbook, relying instead on architecture choices and domain-specific training.
In a separate Fortune interview in February 2026, Hassabis described his longer-term ambition as building toward "radical abundance," a state in which AI has bottled the scientific method and can autonomously run the hypothesis-experiment-publication cycle. "In 10, 15 years' time, we'll be in a kind of new golden era of discovery that is a kind of new renaissance," he told Fortune editor-in-chief Alyson Shontell.
AlphaFold to Isomorphic: From Protein Maps to Drug Candidates
Hassabis has consistently described AlphaFold as the opening chapter rather than the conclusion. The tool, which predicted the 3D structure of approximately 200 million proteins and made those predictions freely available, has been used by more than three million researchers in 190 countries since its public release, according to the Nobel Committee's October 2024 press release. That dataset forms the computational substrate for Isomorphic Labs, the DeepMind spinout Hassabis co-runs, which is now applying AlphaFold 3 to design entirely new small molecules and biologics targeting proteins previously considered undruggable.