Why EV Battery Makers Need to Start Using Machine Learning

Welcome back to The Electric! 

At a time of fast transformation of the transportation system, the lithium-ion battery industry is still largely mired in decades-old ways of invention and development. This week, we look at companies that are using “small data” to create artificial intelligence algorithms and accelerate battery development.

In 2019, Richard Wang, CEO of Cuberg, a developer of lithium-metal batteries, was frustrated by how long it took to design and test electrolytes—the liquid that helps lithium ions flow between the electrodes—before ascertaining whether they would work for his battery. 

Wang thought a machine-learning algorithm could speed things up by identifying the best-known electrolytes. So Cuberg, which is based in San Leandro, Calif., hired Aionics, an artificial intelligence spinoff from nearby Stanford University, to help. Over the subsequent year, the two companies developed a custom model that could rapidly identify promising electrolytes among thousands Wang’s team wanted to test. Despite this progress, the model wasn’t the Goldilocks invention system Wang wanted, one that would find precisely the right electrolyte to make his battery work, saving him years of lab work.

Still, the effort was laudable—and necessary—if Western battery companies and automakers want any chance of competing in an electric vehicle race that’s currently dominated by Tesla and China. While U.S. and European companies are moving rapidly to get their new EVs in line for production by the middle of the decade, they have been much slower to ready their batteries for high-volume manufacture.