By Kyle Proffitt
March 25, 2026 | The Battery Venture, Innovation & Partnering event presentations ended in Orlando this week with two panel discussions, both exploring the challenges inherent in the ever-expanding industries calling for battery support.
The first panel, moderated by Katherine He of TDK Ventures, focused on bridging the gap between GPUs and the grid. The discussion circled back repeatedly to three interlocking challenges: chemistry selection, supply chain concentration, and the timeline mismatch between what data centers need now and what the industry can actually deliver.
Cameron Dales of Peak Energy described Peak’s position producing utility-scale grid storage based on sodium-ion technology and confirmed data center demand as a direct business driver. He estimated that of their contracted off-take of over a million dollars, “maybe 60% of that is directly tied to hyperscale offerings.” He made the case for moving beyond lithium-ion at large scale, citing cost, fire risk, supply chain concentration, and maintenance complexity.
“The core lithium ion technology is probably not the best choice for large grid scale installations … costs are still too high, even with today’s LFP prices, to really reach the true goals of full grid scale storage,” he said. Dales said that “what the hyperscaler data center providers are really providing is reliability; it’s the number one factor on the performance side.” He described Peak’s sodium systems as “fully passive … no moving parts … no cooling systems to break … no need for safety systems to prevent these things from catching fire,” and argued that this simplicity is precisely what makes them suitable for deployment beside multi-billion dollar data centers.
Brad Li of Sineng Electric highlighted his company’s role as a power conversion system provider, noting they had participated in “the world’s largest sodium ion battery system — 100 MW and 200 MWh.” He vilified chemistry tribalism, saying it’s not about “one chemistry over another,” but rather about “system integration and compatibility,” with each part fit for the need. He also championed the approach for a data center of starting off-grid and connecting later, to avoid the multi-year delay in grid interconnects.
Dan Blondal of Nano One Materials lamented our supply chain dependency, noting that for LFP, “anywhere between 95% and 99% of the supply chain is bottlenecked through one nation … and they’ve crushed the price to the point where it’s incredibly difficult for any major refining company, or early stage startup, or even a large company to contemplate investing in the space.” He argued that the solution requires innovation at the precursor and material processing level. “That’s the only place where we can cut the costs; that’s the only way we can decouple from the supply chains,” he said.
David Hynek of Third Derivative/RMI framed data centers as an opportunity that will pull emerging chemistries out of the lab. He reiterated Katherine He’s points that data centers need to cover sub-second power quality applications but also need medium duration and extended duration storage options. He argued that reaching true self-sufficiency would eventually require “100-hour range” storage, a feat accomplished by the Google-Xcel battery installation using iron-air batteries integrated with renewables, and he pointed to other solutions Third Derivative backs, including flow batteries and compressed CO2 storage systems, all gaining ground on the heels of the AI boom.
Drug Discovery vs Battery Complexity Challenge
The second panel, moderated by Anil Achyuta of Energy Impact Partners, focused on AI and hype in the battery industry, specifically how AI might accelerate innovation. A very interesting discussion developed around what a key AI-mediated breakthrough would look like in batteries, such as the discovery of a new material, perhaps a better electrolyte. Participants sparred over a comparison to the world of drug discovery—asking whether batteries, with their many interconnected components, manufacturing challenges … or humans, with their trillion cells and various differences, are more complicated? Achyuta asked, “if the model narrows the search space, but the answer still dies in the synthesis and scale-up, and safety testing and manufacturing, have you really discovered anything commercially useful?”
Qichao Hu of SES AI considered the KPI for AI to be “the dollar per breakthrough,” and said that although the field is in a mixed phase between human and fully AI-driven development, he envisions a future where AI does all of the physical work. “I think within about two to three years, we’ll get to the point where … there’s actually no human in this entire closed workflow.” One human will enter a prompt, and that is the project, he explained. This will send commands to autonomous laboratories that execute experiments with different formulations and produce pristine data. “Then I think we can really minimize dollar per breakthrough,” Hu said.
Venkat Viswanathan, professor at the University of Michigan suggested that batteries are equally as or more complicated than human disease, considering individual idiosyncrasies and all of the challenges of synthesis, scale-up, safety testing, and manufacturing. With regard to drug discovery, he argued that the inflection point has already come, and companies like Isomorphic, of AlphaFold fame, demonstrate that AI-mediated drug discovery is the path forward. While the moment has not yet come in the battery space, he argued that “the fact that it will happen is undeniable.” He added that it is effectively an insurance policy for any large company. “Why would you not make the insurance contract?” he asked.
In Q&A at the end of the panel, an audience member asked who is likely to win in the breakthrough race, when you need abundant data, and nobody wants to share this precious commodity. The audience member suggested that Tesla may be the only player equipped, because they have coordination from cell manufacturer, to automotives, to an AI company.
However, Qichao Hu pushed back, repeating and strengthening his contention that we will create that data fresh, using automated, robotic battery and material testing, floor after floor, lab after lab, collecting data from materials and cells under controlled conditions—no sharing between labs and departments necessary, creating billions of data points for AI, and leading inevitably to the next big breakthrough.
