In short: Speaking on the 20VC podcast with Harry Stebbings in early April 2026, Demis Hassabis described how Google DeepMind has accelerated over the past two to three years by combining the computing resources of Google Brain with DeepMind’s research culture and returning to what he calls a “startup or entrepreneurial” way of working. He also revealed that he was running the group’s pharmaceutical AI spinoff, Isomorphic Labs, as a “second day of work” before expected human trials in oncology later this year, starting around 10 p.m.
Collection of ingredients
Google DeepMind’s official merger of DeepMind and Google Brain was completed in 2023. Hassabis has since described it as a period of deliberate acceleration: “aligning talent within the company, kind of pushing in one direction,” gaining access to what DeepMind previously called “scale and unfocused” computing infrastructure. According to his characterization, the transformation required as much cultural adjustment as structural: the organization “had to almost go back to our startup or entrepreneurial roots and be nimbler, be faster, ship things really quickly.” He said the current competitive environment was “ferocious”. Veteran workers with 20- and 30-year careers told him it was “the most intense environment they’ve ever seen, maybe ever in the tech industry.”
Hassabis said he talks to Alphabet CEO Sundar Pichai “every day,” reflecting the extent to which Google DeepMind is now at the operational center of Alphabet’s product and research strategy. This proximity is matched by a correspondingly large capital commitment. Google’s computing architecture, developed in part through custom chip partnerships with companies including Broadcom, is at the heart of this deployment.: Alphabet has spent $91.4 billion in capital expenditures in 2025 and is targeting between $175 billion and $185 billion in 2026, doubled by supply constraints rather than capital availability being described as the main limiting factor.
90% claims
One of Hassabis’ more ambitious statements on the podcast was about DeepMind’s contribution to the history of artificial intelligence. According to him, about 90% of the breakthroughs that laid the foundations of the modern artificial intelligence industry were produced by Google Brain, Google Research or DeepMind. The claim is broadly consistent with academic notes on foundational developments, including the transformer architecture produced by Google Brain in 2017, early work on reinforcement learning from human feedback, and deep reinforcement learning techniques developed at DeepMind. The 2024 Nobel Prize in Chemistry was awarded to Hassabis and John Jumper and shared with David Baker for the AlphaFold protein folding system. Whether 90% is accurate as a proportion is a matter of interpretation, and the industry has proliferated significantly since these seminal papers. The frame functions as a position statement as much as a historical claim.
The operational result of this legacy is a dramatically accelerated product release cadence. Google’s open weight model program, most recently Gemma 4, is now releasing models built from the same research and training infrastructure as Gemini 3.bridging the gap between frontier research and pre-existing open source contributions. Gemini reached approximately 750 million monthly active users by the end of the fourth quarter of 2025, Gemini 3 was described in a secondary report released in November of that year as prompting an immediate internal response at OpenAI.
Second working day
In addition to leading Google DeepMind, Hassabis also runs Isomorphic Labs, DeepMind’s 2021 pharmaceutical AI arm. He described his routine in a 20VC chat: his first day at DeepMind, followed by “second working day” dedicated to Isomorphic’s drug discovery program beginning around 10 p.m.The dual commitment reflects the belief that applying AI to drug discovery is both Hassabis’ most important long-term ambition and a project that requires continued personal involvement rather than delegation.
Isomorphic raised $600 million in April 2025 and has existing partnership agreements with Eli Lilly and Novartis with combined milestone values of up to $3 billion. In February 2026, the company released IsoDDE, a drug design tool that Isomorphic says doubles the accuracy of AlphaFold 3 for generating drug candidates. Human clinical trials in oncology are expected at the end of 2026. Competitive dynamics in AI-driven drug discovery are intensifying across the industry: Anthropic’s April 2026 acquisition of Coefficient Bio, a stealth startup founded by former Genentech computational biology researchers, for about $400 million.signals that general-purpose AI companies are now treating pharmaceutical discovery as a product category rather than just a demonstration of model capability.
Competitive framework
Sebastian Mallaby’s 20VC podcast talk as a biography of Hassabis, “Infinity Machine”, published on March 31, 2026 and based on more than 30 hours of interviews, presents a researcher entering the most commercially relevant phase of his career with a consistent thesis: that the most important research can be done both separately and simultaneously, the most important products are both separate organizations and the most important products. will determine the shape of the industry across borders. 2025 has consolidated AI as a central strategic priority in the technology industrywith its capital, talent and institutional structure all restructured around the question of pace. For Hassabis, the answer has been to bring the speed of a startup to the resource base of one of the world’s largest technology companies, and view the combination as a lasting advantage.
The scale of capital flowing into the field makes it difficult to sustain this advantage. SoftBank’s $40 billion bridge loan to OpenAI it even represents a form of capitalization that Alphabet’s computing liabilities cannot trivially overlap in kind. Hassabis’s account of a “ferocious” competitive environment is not rhetorical: it is a structural description of a race in which the resources of incumbents and the ambitions of rivals converge to such a point that institutional inertia is not merely a disadvantage but a disqualification. The startup mentality he describes at Google DeepMind is more of a necessity than an advantage in this context.
