Revolutionizing Material Science: DeepMind’s AI Discovers Over 2 Million New Crystals


DeepMind, a leading AI research lab, has made a groundbreaking discovery in the field of material science. Using their new deep learning tool, Graph Networks for Materials Exploration (GNoME), they have identified 2.2 million new crystal structures, including 380,000 stable materials with potential applications in various future technologies.

Key Takeaways:

  • DeepMind’s AI tool, GNoME, discovers 2.2 million new crystals.
  • Among these, 380,000 are identified as stable, viable for future technologies.
  • The discovery could impact areas like superconductors, batteries, and solar panels.
  • GNoME’s predictions are now available to the global research community.

A Leap in Material Discovery

Traditionally, discovering new materials has been a slow and labor-intensive process, often involving months of experimentation. However, GNoME has changed the game by predicting the stability of new materials, thereby accelerating the discovery process. This AI-driven approach has expanded the number of known stable materials from 48,000 to a staggering 421,000.

Impact on Future Technologies

The materials discovered by GNoME hold immense potential for developing transformative technologies. For instance, among the new materials are 52,000 layered compounds similar to graphene, which could revolutionize electronics with the development of superconductors. Additionally, 528 potential lithium-ion conductors were identified, which could significantly enhance rechargeable battery performance.

Methodology Behind the Discovery

GNoME utilizes two pipelines for discovering stable materials: a structural pipeline that creates candidates with structures similar to known crystals, and a compositional pipeline that follows a more randomized approach. The predictions are then evaluated using Density Functional Theory calculations, a standard in assessing material stability.

Collaborations and Real-World Applications

The practicality of GNoME’s predictions has been validated by external researchers, with 736 of the new materials already created experimentally in labs worldwide. Furthermore, in collaboration with Lawrence Berkeley National Laboratory, a second paper demonstrates how AI predictions can aid in autonomous material synthesis.

The Future of Material Science

This discovery by DeepMind is not just a milestone in AI and material science but also a beacon for future research and development. By making these predictions accessible, DeepMind hopes to drive forward research in inorganic crystals and leverage machine learning tools for experimental guidance.


DeepMind’s breakthrough in discovering millions of new materials through AI marks a significant advancement in material science. This development not only accelerates the discovery of new materials but also opens up new possibilities for technological advancements, potentially leading to more sustainable and efficient future technologies.

Jonathan Browne
Jonathan Browne
Jonathan Browne is the CEO and Founder of Livy.AI

Read more

More News