Microsoft set AI loose on over 32 million candidate materials to find a new one that could cut lithium in batteries by up to 70%.
- Microsoft showed AI screening of 32+ million candidate materials for batteries (July 14, 2025).
- The result: a new material that could cut lithium in batteries by up to 70%.
- Behind it: MatterGen (generates candidates) and Aurora (foundation model for atmosphere, land and oceans).
32 million candidate materials. No engineer sits down and goes through them by hand. Microsoft set AI loose on exactly that - and it pulled out one that could cut lithium in batteries by up to 70%. The work ran through two of their models - MatterGen for new materials and Aurora for earth sciences.
The core of it is the scale of the screening - a person can't go through 32 million variants, AI can. It doesn't invent the material in the scientist's place. It hands them the short list to work from.
Wait and see if it holds: cut the dependence on lithium, and the whole battery chain - price, supply, geopolitics - gets lighter. But honestly, these materials are still in the lab. Actual production is years out.
The findings are at lab stage; the road to actual production is several years out.