Demis Hassabis stated that Google DeepMind needed to revert to its startup origins following the Brain merger.

      In summary: Demis Hassabis, during an early April 2026 conversation on the 20VC podcast with Harry Stebbings, elaborated on how Google DeepMind has increased its pace over the last two to three years by integrating Google Brain’s computational capabilities with DeepMind’s research ethos, returning to what he referred to as a “startup or entrepreneurial” approach. He also mentioned that he manages Isomorphic Labs, the group's pharmaceutical AI offshoot, during a “second workday” starting around 10 PM, in anticipation of upcoming human trials in oncology later this year.

      Gathering the essentials

      The official merger of Google DeepMind and Google Brain was finalized in 2023. Hassabis characterized the time since as a period of intentional acceleration, aligning talent “from around the company, pushing in one direction,” accessing much-needed computational infrastructure, and fostering what he termed a “relentless sort of focus and pace.” According to him, this transformation necessitated both cultural and structural changes: the organization needed to revert to its “startup or entrepreneurial roots” while being more agile and quick to deliver results. He described the current competitive landscape as “ferocious,” with veteran employees claiming it is “the most intense environment they’ve ever witnessed, perhaps in the entire technology industry.”

      Hassabis indicated that he communicates with Sundar Pichai, Alphabet’s CEO, “every day,” highlighting the extent to which Google DeepMind now plays a central role in Alphabet’s product and research strategy. This connection is further supported by significant capital investment. Google’s expansion in computing, partly driven by its partnerships with custom chip manufacturers like Broadcom, is a critical aspect of this positioning: Alphabet allocated $91.4 billion in capital expenditures in 2025, with projections of $175 billion to $185 billion for 2026, nearly doubling, primarily limited by supply constraints rather than funding availability.

      The 90% assertion

      One of Hassabis’s more bold claims during the podcast was regarding DeepMind’s impact on the history of AI. He stated that about 90% of the innovations that form the foundation of the modern AI industry were developed either by Google Brain, Google Research, or DeepMind. This statement generally aligns with the academic records of foundational advancements, such as the transformer architecture created by Google Brain in 2017, early work on reinforcement learning through human feedback, and deep reinforcement learning methods established at DeepMind. The 2024 Nobel Prize in Chemistry awarded to Hassabis, John Jumper, and David Baker for the AlphaFold protein-folding system stands as the most formally recognized achievement among these. Whether the 90% figure is accurate is subject to interpretation, and the industry has significantly diversified since those early research papers. The framing serves both as a positioning statement and a historical reference.

      An operational result of this legacy is a sharply accelerated product release schedule. Google’s open-weight model initiative, most recently Gemma 4, now disseminates models developed using the same research and training infrastructure as Gemini 3, bridging a gap between leading-edge research and open-source contributions that had previously existed. By the end of the fourth quarter of 2025, Gemini had approximately 750 million monthly active users, with Gemini 3 reportedly eliciting an urgent internal response from OpenAI upon its release in November of that year.

      The second workday

      In addition to leading Google DeepMind, Hassabis also oversees Isomorphic Labs, the pharmaceutical AI spin-off that DeepMind launched in 2021. He discussed his working arrangement in the 20VC conversation, which consists of a first workday at DeepMind, followed by a “second workday” starting around 10 PM focused on Isomorphic’s drug discovery initiatives. This dual commitment reflects his belief that applying AI to drug discovery is both his most significant long-term goal and a project requiring his ongoing personal involvement rather than delegation.

      Isomorphic secured $600 million in April 2025 and has established partnership agreements with Eli Lilly and Novartis, with combined milestone values reaching up to $3 billion. In February 2026, the firm released IsoDDE, a drug design tool that Isomorphic claims enhances the accuracy of AlphaFold 3 for generating drug candidates. Human clinical trials in oncology are anticipated later in 2026. The competitive landscape for AI-driven drug discovery is intensifying across the sector: Anthropic’s acquisition of Coefficient Bio for roughly $400 million in April 2026—a stealth startup founded by former Genentech computational biology researchers—indicates that general-purpose AI firms are increasingly viewing pharmaceutical discovery as a product category, rather than simply showcasing model capabilities.

      The competitive context

      The discussion on the 20VC podcast, similar to Sebastian Mallaby’s biography of Hassabis, “The Infinity Machine,” published on March 31, 2026, and based on over 30 hours of interviews, portrays a researcher at the peak of his commercially urgent phase, maintaining a consistent thesis: