DeepMind to Begin AI-Driven Cancer Drug Trials
The CEO of Google DeepMind has announced that clinical trials for cancer drugs designed by artificial intelligence will begin this year. The company also expects breakthroughs in robotics within the next 18 months, positioning AI as a direct driver of medical and physical automation innovation.
- The drug discovery efforts are led by Isomorphic Labs, a company spun out of DeepMind in 2021 to leverage its AI research for creating new medicines. The underlying technology, AlphaFold, earned its developers a Nobel Prize in Chemistry in 2024 for its ability to accurately predict the 3D structure of proteins. - The latest version of the AI, AlphaFold 3, goes beyond just protein structure to predict how proteins interact with other molecules like DNA, RNA, and ligands, which is a critical step in designing new drugs. This AI is the first to surpass physics-based tools for predicting biomolecular structures. - Isomorphic Labs is not working alone; it has established major partnerships with pharmaceutical giants Novartis and Eli Lilly, with deals valued at nearly $3 billion. The company also secured $600 million in its first external funding round in March 2025. - While the first clinical trials are expected in early 2026, Isomorphic Labs currently has 17 drug discovery projects in its pipeline. The initial therapeutic focus for the upcoming human trials will be oncology. - The global market for AI in drug discovery was valued at over $2.3 billion in 2025 and is projected to exceed $13 billion by 2033, with oncology being the largest segment. The traditional drug development process takes an average of 10 years with only a 10% success rate. - In a related project with Yale University, DeepMind developed a separate AI model named C2S-Scale 27B which analyzed over 4,000 drugs to find compounds that could make cancer cells more visible to the human immune system. - The referenced breakthroughs in robotics are based on models like Gemini Robotics, which are designed to give robots a more nuanced understanding of their environment and allow them to follow complex, multi-step commands.