Demis Hassabis mentioned that Google DeepMind needed to revisit its startup origins following the Brain merger.
In summary: On the 20VC podcast with Harry Stebbings in early April 2026, Demis Hassabis shared how Google DeepMind has sped up its progress in the last two to three years by combining Google Brain’s computational resources with DeepMind’s research ethos, returning to a more "startup or entrepreneurial" approach. He also mentioned that he manages Isomorphic Labs, the pharmaceutical AI spinoff of the group, during what he calls his "second workday" starting around 10 PM, in preparation for anticipated human trials in oncology later this year.
Gathering the elements
The formal integration of DeepMind and Google Brain into Google DeepMind was finalized in 2023. Hassabis characterized the subsequent period as one of intended acceleration, uniting talent from across the organization and leveraging computational infrastructure that was previously not available at scale for DeepMind. He emphasized what he referred to as a “relentless sort of focus and pace.” According to him, the transformation necessitated cultural shifts as much as structural ones, urging the organization to reconnect with its "startup or entrepreneurial roots" to be more agile and deliver quickly. He described the current competitive landscape as "ferocious," with long-tenured employees claiming it's the most intense environment they've ever encountered in the tech industry.
Hassabis noted that he communicates with Sundar Pichai, Alphabet’s CEO, "every day," indicating the extent to which Google DeepMind is now central to Alphabet’s product and research strategy. This close relationship is supported by significant capital investment. Google’s infrastructure expansion, partly developed through custom chip collaborations with firms like Broadcom, plays a crucial role: Alphabet allocated $91.4 billion for capital expenditures in 2025 and projected between $175 billion and $185 billion for 2026, nearly doubling the figure, with supply constraints cited as the main limiting factor.
The 90% assertion
One of Hassabis’s more bold statements during the podcast was regarding DeepMind’s role in the evolution of AI. He claimed that around 90% of the breakthroughs that underpin the modern AI sector originated from either Google Brain, Google Research, or DeepMind. This assertion generally aligns with academic records on major innovations, such as the transformer architecture from Google Brain in 2017, initial efforts on reinforcement learning from human feedback, and deep reinforcement learning methods developed at DeepMind. The 2024 Nobel Prize in Chemistry, awarded to Hassabis and John Jumper and shared with David Baker for the AlphaFold protein-folding system, stands out as the most formally recognized of these accomplishments. Whether the 90% claim accurately reflects the proportions is open to interpretation, especially considering the diversification in the industry since those foundational studies were published. The framing serves as much as a positioning statement as it does a historical assertion.
The operational impact of this legacy has resulted in a significantly increased product release pace. Google’s open-weight model program, with its latest release Gemma 4, now presents models developed using the same research and training framework as Gemini 3, bridging the gap that once existed between cutting-edge research and open-source contributions. By the end of the fourth quarter of 2025, Gemini had around 750 million monthly active users, with reports indicating that Gemini 3 elicited an urgent internal response at OpenAI upon its launch in November 2025.
The second work period
In conjunction with his leadership role at Google DeepMind, Hassabis oversees Isomorphic Labs, the pharmaceutical AI spinoff established by DeepMind in 2021. He outlined his working schedule during the 20VC discussion: a primary workday at DeepMind followed by a "second workday" starting around 10 PM focused on Isomorphic’s drug discovery efforts. This dual commitment underscores his belief that integrating AI into drug discovery is both a critical long-term objective and an initiative requiring his ongoing direct involvement rather than delegation.
Isomorphic secured $600 million in funding in April 2025 and has partnerships with Eli Lilly and Novartis with potential milestone values reaching $3 billion. In February 2026, the company launched IsoDDE, a drug design tool that reportedly enhances AlphaFold 3's accuracy in generating drug candidates by twofold. Human clinical trials in oncology are anticipated later in 2026. The competitive landscape for AI-driven drug discovery is escalating, evidenced by Anthropic’s acquisition of Coefficient Bio for around $400 million in April 2026, a stealth startup formed by former Genentech computational biology experts, indicating that general-purpose AI firms are now viewing pharmaceutical discovery as a market category rather than merely a demonstration of model capabilities.
The competitive context
The 20VC podcast discussion, similar to Sebastian Mallaby's biography of Hassabis, "The Infinity Machine," released on March 31, 2026, and based on over 30 hours of interviews, showcases a researcher who has entered the most commercially pressing chapter of his career with a unified theme: that
Other articles
Demis Hassabis mentioned that Google DeepMind needed to revisit its startup origins following the Brain merger.
Hassabis mentions that Google DeepMind needed to revisit its startup origins and adopt a more resourceful approach following the Brain merger, managing Isomorphic Labs as his side job starting at 10 PM.
