The data center boom could shift priorities for battery makers

Argonne National Laboratory is best known for its nuclear research — after all, it was originally established as part of the Manhattan Project. But the lab, located in Illinois, also has a long history of energy storage research and development, and its focus on batteries for both transport and stationary storage is one of its defining features today.

Venkat Srinivasan, who leads Argonne’s energy storage programs, said that both applications face overlapping challenges, but scaling grid-scale stationary storage is uniquely complicated — and its trajectory to scale may be determined in part by the shape of the data center boom.

“When it comes to batteries, everything is a compromise,” Srinivasan said. “There are multiple metrics you’re trying to satisfy simultaneously, and you can't maximize everything. You’re asking yourself, what vector am I going to really maximize, and which others can I just manage?”

Battery makers and their customers are always attempting to balance considerations like cost, lifespan, and safety, he said. For stationary storage, for instance, low cost has traditionally been the top priority, largely because of the price sensitivity of end consumers of electricity. But the scale and pace of the data center boom is complicating that expectation — though Srinivasan said that the extent of its impact on the storage industry may not be felt until the dust of the AI race settles.

The battery question

Unlike the EV industry, the energy sector can’t justify higher-cost products just because they might be more appealing to a certain class of customers, Srinivasan said. That’s because the energy sector is essentially just selling electrons, no matter how they’re packaged — and of course there’s the concern that adding expensive tech to the grid could increase electricity costs for consumers.

“It’s not like an electric car where we’re asking people to substitute an existing vehicle with something we think is better for the environment, better for the driving experience,” he said. “We have to provide them with reliable electricity and it has to be at parity with the existing price.”

That cost vector is also more complex with stationary storage because of the extremely varied applications: shorter versus longer durations, for example.

Today the market is dominated by lithium-ion batteries in the four-hour range, but that chemistry may even make economic sense for durations of up to around 10 or 12 hours, Srinivasan said. For those batteries, developers and utilities may have an easier time making up the upfront cost over time because they’re generally used more often.

But as the duration gets longer, the less the storage gets used, he added. Events where those batteries need to be discharged for the duration they’re designed for should, in theory, be extremely rare.

“You’re building for resiliency in those cases, and so the technology is not going to be used that much, and the upfront cost better be low in order for it to ultimately be low cost to the consumer,” he said. ”You can’t tell consumers that on the week they might lose electricity it’s going to cost 100 times what it normally costs.”

That assessment is at the core of why lithium-ion batteries won’t be the winners for grid-scale storage in the long-term: as duration goes up, cost has to come down. And the materials and manufacturing cost of lithium batteries can’t sustain that, Srinivasan said.

The AI twist

But given both the increasing urgency to firm up renewables and support the grid with storage, and the massive budgets of the world’s top AI companies, it’s possible that cost won’t remain the central measure of success for much longer, Srinivasan said.

When it comes to data center developers and operators, including the hyperscalers, the most important vector is reliability. Audrey Lee, Microsoft’s director of energy strategy, said at a conference earlier this summer that the company’s need for 99.999% uptime means it isn’t willing to “go first when it comes to deployment of a new technology and that technology risk.”

In those cases, keeping costs low may well become the priority that buyers are willing to forgo, said Srinivasan.

“The question becomes, as the data center boom comes up, is where that money is going to be made, and how much money [companies] are willing to spend on having resilient infrastructure that is also carbon free,” he said.

With AI, there’s the assumption that companies will be able to make up for the high cost of storage elsewhere, but exactly where those gains will come from isn’t yet clear: 

“We’re still in this loop of throwing a lot of money at a lot of different things…so it’s hard to know where that particular market is going to settle, and the battery equilibrium may not show up until they figure that out,” Srinivasan said.