- EV sales have been slower than expected recently as consumers worry about cost, charging and range.
- Startups and scientists are trying to alleviate these concerns by using AI to design EV batteries.
- Experts say AI could dramatically speed up battery development, bringing us closer to better EVs.
EV sales appear to be stagnating.
Demand for vehicles has cooled in the United States over the past year, with slower-than-expected sales growth causing some automakers to pause big investments.
Surveys show that consumer concerns about EVs include cost, charging, range, etc. All of this stems from the most expensive and critical component of an electric vehicle: the battery.
Jason Kohler, co-founder of battery startup Chemix, thinks that could change thanks to AI.
“When you think about EV adoption, almost every reason people don't want to buy an EV has to do with the batteries,” he told Business Insider.
“Whether it's too expensive, the range is too short, the charging speed isn't fast enough or there are safety concerns, it all comes down to the battery,” he added.
The wave of AI hits EVs
California-based Chemix, which raised $20 million in investment in April, is one of several startups and research institutions looking to solve many of these problems with the help of AI.
Chemix is using machine learning algorithms to develop EV batteries that charge faster, hold more energy, and last longer than current EV power units.
Another common complaint drivers have about EVs is that they tend to perform poorly in hot and cold temperatures, an issue that Chemix's custom-designed batteries could help solve.
The startup's AI technology can also be used to filter out common battery materials such as nickel and cobalt, which have been linked to human rights abuses.
Kohler said a major benefit of using AI in EV battery design is that it dramatically speeds up the process: A 2020 study published in Nature found that a machine learning model could reduce the time it takes to identify a fast-charging battery design from 500 days to 16 days.
“The problem we're trying to solve is to increase the pace of battery development,” Koller said, adding that this is crucial to producing electric vehicles that can compete with those with internal combustion engines.
“There's no question that we need to make significant advances in battery performance, and we believe that getting there will require a fundamentally different approach to battery development,” he added.
Battery Boost
Micah Ziegler, an assistant professor at the Georgia Institute of Technology, told BI that designing an EV battery is a three-step process.
Before scientists can develop recipes for synthesizing battery materials and test them in the lab, they must first determine the structure and combination of elements that will produce battery materials with the desired properties.
Ziegler said the variety of components needed for an EV battery, along with the various requirements for lifespan, safety, cost and charging speed, makes battery design a long and difficult process.
“The number of options is huge. When you start combining different elements, you can end up with trillions of combinations,” he added.
AI can dramatically speed up this process, identifying promising combinations much faster than humans could.
Karl Mueller, a physical chemist at Pacific Northwest National Laboratory, describes battery design as an “Edisonian” effort — a lengthy trial-and-error process in which scientists compare and fine-tune billions of chemical combinations.
“It's a slow process and it can take years to find and improve new materials,” he told BI.
PNNL recently partnered with Microsoft to use the company's AI and cloud computing technology to discover promising new materials for making EV batteries.
The solid-state electrolyte they discovered is less reliant on lithium, an increasingly scarce resource that's at the heart of today's EV batteries, and is also less likely to catch fire than lithium-ion batteries.
“It's going to be phenomenal.”
Mueller said incorporating AI into PNNL's workflow significantly sped up the process of narrowing down 32 million chemical combinations to about 20 battery designs.
“The latest AI tools we're developing with Microsoft will enable us to design for specific characteristics and quickly eliminate those that don't appear to have those characteristics,” he said.
“I think this will do wonders for speeding up the discovery process, speeding up discovery of what we need to do to get to more electrification and better electric vehicles,” Mueller added.
The hype around generative AI may be dying down in some quarters, but one area is already seeing clear benefits: the scientific community.
Models such as DeepMind's AlphaFold 3 could improve mapping of complex biological proteins, creating new opportunities for researchers to rapidly develop life-saving medicines.
Mueller said the collaboration between PNNL and Microsoft has convinced him that generative AI has applications far beyond speeding up EV battery design.
“The idea of building a scientific agent based on a large-scale language model and then connecting that to a discovery workflow is really exciting,” he said.
“I think in a year or two we won't know how far the science has come,” he added.
