What Happened to DeepMind's AlphaFold?

DeepMind's AlphaFold emerged as a groundbreaking artificial intelligence program designed to predict the three-dimensional structures of proteins, a challenge known as the 'protein folding problem' that had perplexed scientists for over 50 years. Its initial version, AlphaFold 1, demonstrated remarkable accuracy at the Critical Assessment of Structure Prediction (CASP) in 2018, outperforming conventional methods. The true breakthrough came with AlphaFold 2 in 2020, which achieved a level of accuracy comparable to experimental methods, effectively solving the protein folding problem for single proteins. This achievement was widely hailed as transformational, significantly accelerating research in biology and medicine.

The impact of AlphaFold 2 was rapidly amplified through its open-sourcing in July 2021, alongside the launch of the AlphaFold Protein Structure Database (AFDB) in partnership with EMBL-EBI. This database, by July 2022, contained predictions for over 200 million protein structures from 1 million species, making nearly all known proteins on the planet freely accessible to researchers. This unprecedented access to structural information has been credited with saving millions of dollars and hundreds of millions of years in research time, fostering discoveries in areas like malaria vaccines, cancer treatments, and enzyme design.

A key turning point for commercial application was the establishment of Isomorphic Labs in November 2021, an Alphabet subsidiary spun out of DeepMind with the mission to leverage AI for drug discovery. This move signaled Google's intent to commercialize the immense potential of AlphaFold's technology. In May 2024, Google DeepMind and Isomorphic Labs jointly announced AlphaFold 3, a revolutionary model that expanded prediction capabilities beyond proteins to include DNA, RNA, ligands, and ions, and their complex interactions. AlphaFold 3 demonstrated at least a 50% improvement in accuracy for protein-molecule interactions compared to existing methods, further enhancing its utility for drug design. The AlphaFold Server was also launched, providing free access to AlphaFold 3's capabilities for non-commercial research.

The profound scientific contributions of AlphaFold were recognized with the 2024 Nobel Prize in Chemistry, awarded to Demis Hassabis and John Jumper for their work on protein structure prediction, shared with David Baker for computational protein design. The source code and weights for AlphaFold 3 were made available for non-commercial use in November 2024 and publicly in February 2025, further democratizing access to this powerful tool for academic research.

However, as of July 29, 2026, a significant strategic shift has occurred within Google DeepMind. The dedicated AlphaFold research team has been disbanded, with most of its original authors and key members either reassigned internally to projects focused on Google's flagship Gemini large language model, nuclear fusion, genomics, and enzyme design, or departing the company entirely, with some joining competitors like Anthropic. This restructuring reflects DeepMind's evolving strategy to move away from isolated teams tackling single scientific grand challenges towards developing more general-purpose AI systems capable of assisting scientists across various fields and automating parts of the research process. Despite the team's dissolution, the AlphaFold technology, including AlphaFold 3 and the AlphaFold Protein Structure Database, remains available and continues to be a foundational tool for global scientific research and drug discovery, with Isomorphic Labs actively applying it in commercial pharmaceutical partnerships.