GNoME predicts 2.2 million new crystals, 380,000 stable, but novelty and usefulness are disputed
Disputed
Importance: major (4 of 5)The takeaway
DeepMind’s GNoME (Nature, Nov 2023) used graph neural networks and active learning with DFT to predict 2.2 million new inorganic crystal structures, 380,000 of them computed to be stable.
Status
- Claim
Disputed
- Our reporting
- High confidence
- Verification
- Peer-reviewed
- Importance
- Major (4 of 5)
- Last verified
- 16 October 2026
Your AI and this story
- GPT-6 AstraIn its training data
- Claude Opus 5.5In its training data
- Gemini 4 HydrogenIn its training data
- Grok 4.7In its training data
It happened before the cutoffs of all four assistants, so it can be in their training data.
Key facts
- 2.2M new structures; 380k predicted stable; ~400k added to the Materials Project
- DeepMind: over 700 had already been independently synthesised by other groups
- Cheetham & Seshadri (Chem. Mater., Apr 2024): ‘scant evidence for compounds that fulfill the trifecta of novelty, credibility, and utility’
- Oct 14, 2026 (C&EN, Dalmeet Singh Chawla): new outside research says the GNoME database is ‘littered with duplicates of known and hypothesized crystal structures found in other databases’ and contains unrealistic compounds; Google DeepMind did not comment (article partly paywalled; the study’s authors and figures were not read)
- GNoME lead Ekin Doğuş Çubuk later co-founded Periodic Labs (2025)
What happened
DeepMind scaled ML-guided materials screening by orders of magnitude and released the predicted structures to researchers.
Why it matters
It is the largest AI materials-prediction effort, and a leading example of the gap between computational “discovery” and a useful new material.
Sources
4 sources from 4 sites. Numbers match the chips in the text.
4 sources: 2 primary, 1 press, 1 reaction
Primary
- Scaling deep learning for materials discovery (Nature)nature.com, paper
- DeepMind: Millions of new materials discovered with deep learningdeepmind.google, official
Press
- C&EN: Google DeepMind AI crystallography database includes near duplicate structures and unrealistic compounds (Oct 14, 2026; partly paywalled)cen.acs.org, press
Reactions
- Cheetham & Seshadri critique (Chemistry of Materials)pubs.acs.org, discussion
Changes
- C&EN report (Oct 14, 2026) on near-duplicate and unrealistic structures in the GNoME database
- Linked Periodic Labs entry
- Filed