Software’s Moment: As Battery Technologies Mature, AI Steps In

By Allison Proffitt

June 24, 2024 | The battery market is about ten years old, and while in-house software solutions have carried battery R&D so far, the market is now at a tipping point for new software innovations that will change lithium ion battery research and applications, contends Tyler Lancaster, partner at VC firm Energize Capital.

Energize Capital manages about $1.4 billion in investments targeted at digital platforms to accelerate climate solutions. Lancaster—an environmental engineer by training—leads a lot of the firm’s work on electrification and renewables. Initially that meant solar, but within the past few years his attention has been increasingly captured by battery technologies and the market growing around them.

The first wave of battery software companies was launching about the same time as Energize, in 2016, he says. “Unfortunately, most of them were too early and, frankly, did not work out.” But in the past two to three years, there’s been an inflection point, he says, both in terms of decreasing cost per kilowatt hour and actual unit deployment of batteries.

“Our fundamental position is that because batteries are so cheap and we’ve solved problems manually with people power to get to this point of the industry’s maturity, we really think software and AI/machine learning data is what helps unlock the next phase of scale for the industry.”

Just in the past year, Lancaster observes, battery firms have started to adopt software in a way that wasn’t happening three or four years ago.

Following Solar’s Model

Lancaster compares the battery market to the solar energy market from about 15 years ago. A leading technology emerged with the right balance of production efficiency and economies of scale for manufacturing, which commoditized the hardware. “Crystalline silicon kind of beat out all the competing solutions and that in turn unlocked a whole host of new business model opportunities,” he says. “We’re getting more of this technology in consumers’ and businesses’ hands, but now the big challenge is something is what we call ‘soft costs’: everything not related to the hardware itself.” Lancaster includes marketing, engineering and design work, permitting, and more among those soft costs. He estimates that in the United States, 60% to 70% of a solar array’s cost is now soft costs, not the equipment.

“We think the exact same thing is happening in the battery space, but with the lithium-ion platform and derivative chemistries of it,” he says. The batteries themselves have, “become massively commoditized as China has scaled up production,” but as hardware costs go down, soft costs are rising. Lancaster sees the trend across various battery markets: electric vehicles, grid scale applications, behind the meter applications, even consumer electronics. “We think there’s a whole host of opportunities for software to play a role in reducing those soft costs and really streamlining access for the folks who are adopting batteries.”

Design and Deployment

The two broad areas that Energize sees as ripe for software growth—though certainly not the only ones—are battery research and development (including materials science and chemistry, and in-life analytics monitoring battery degradation and thermal runaway risk) and economic optimization of batteries.

Battery design—even if the chemistry is limited to lithium ion—is a “multivariate decision-making process”, Lancaster says. Battery manufacturers are beginning to leverage AI to simulate different combinations of everything from materials to pack design to cell chemistry for different applications.

For example, a battery designed for a Porsche requiring high torque will be different from a battery designed for a long-range vehicle, he says. AI and ML software can help model how batteries will behave in different use cases to optimize for the right features, he says. The same software can gather battery performance information from an electric vehicle’s telemetry system or an inverter in a grid-scale scale storage set up and use those data to further train the model.

“One of our portfolio companies—Twaice—tackles both of these use cases, both the upfront design piece, and then on an ongoing basis, they call it in life analytics, taking that operating data and then using that to recalibrate and make their upfront design assimilation accuracy even better,” Lancaster says.

Power Market Optimization

But modeling and simulation software is not only useful in product design and predicting lifespan and best use cases. Lancaster points out a significant role for software designed to optimize the ongoing economics of owning and using a battery—particularly in the energy grid where batteries are opening up new use cases.

“For example,” he says, “a battery could do what’s called energy arbitrage and charge up when prices are cheap and then discharge when prices are high. But they also could get paid for simply being available. We have this concept of capacity in our energy markets, which it means you are as a electricity production facility, you are paid simply to be available, uh, to be called upon by the grid operator when conditions on the grid get tight and those are very complex trade-off decisions that you need to make understanding what’s the revenue potential of these different products. How will it actually impact the battery if I dispatch the battery for different use cases in terms of degradation?”

This is a perfect use case for modeling software, Lancaster says, and highlights that the battery market is only just now mature enough to do this sort of strategic optimization.
“The optimization model will need to be like highly tuned because in different geographies, the market rules are different. The products available are different. Operating conditions are different.”

Batteries have unique dispatch characteristics compared to other energy sources—gas or coal or nuclear plants—so understanding the best way to use battery resources in light of various market needs is complex and new.

What’s Next

Energize expects there to be multiple software solutions built to meet different subsets of these problems. And the market is not finished maturing. Lancaster also predicts further opportunities in managing battery end-of-life and second life.

“We’re just early in the cycle for that, just because frankly, many of these batteries have been deployed in the last six, seven years,” he explains. But soon he foresees software tools for certifying how much capacity is left in batteries for second life uses, for example, or optimizing recycling.

Lancaster also predicts increasing partnerships between software development firms and OEMs to share data to train foundation models. “It’s a symbiotic relationship because those [OEM] firms don’t have access to the type of best of breed software development approaches; they don’t have access to the battery engineers.”