The power grid was not built for what is being asked of it right now.
AI data centers are driving the most significant surge in electricity demand in modern U.S. history, and the infrastructure designed to deliver that power is struggling to keep pace. BloombergNEF projects that peak demand from data centers alone could reach 106 GW by 2035, representing roughly 8.6% of all U.S. electricity demand.
For Brooks Sherman, a strategy and business development professional who researched next-generation battery technologies for his MBA capstone at the University of Vermont’s Grossman School of Business, this bottleneck is doing more than straining the grid. It’s creating one of the strongest demand signals the battery storage industry has ever seen. “Most of the conversation about AI infrastructure is focused on compute power and chip supply,” says Sherman, whose capstone, Saving for the Future: Strategy in a Post-Lithium Energy Storage Market, examined grid-scale and distributed battery technologies. “But the energy question is catching up fast. You can’t run a 500-megawatt data center campus on a grid connection that takes seven years to build. That mismatch is forcing the entire industry to rethink how power gets delivered, and battery storage is at the center of that rethink.”
A Grid Under Pressure
The numbers tell a stark story:
- In Texas, grid operator ERCOT was tracking roughly 226 GW of large load interconnection requests by late 2025, with data centers accounting for more than 70% of that pipeline.
- In the PJM region, which covers 13 states in the mid-Atlantic and Midwest, load forecasts project 30 GW of new data center demand through 2030, which is roughly 94% of PJM’s total projected load growth over that period.
- The timeline from initial interconnection application to energization runs five to seven years in most cases. In congested areas near urban centers, it can stretch to ten.
The economic pressure is compounding. Capacity auction prices in PJM markets climbed sharply between 2024 and 2026 as demand forecasts rose and new generation capacity lagged behind. The International Energy Agency projects global data center electricity consumption will climb from roughly 415 TWh in 2024 to around 945 TWh by 2030, driven primarily by AI workloads that can require significantly more power per rack than traditional cloud computing. These aren’t projections about a distant future but concrete data points that describe conditions already shaping investment decisions and site selection for data center developers today.
Why Storage Is the Missing Piece
Grid constraints are pushing data center developers toward on-site power generation, which are often natural gas-fired systems that can be deployed in 18 to 24 months instead of waiting years for a grid connection. But every on-site power installation, whether isolated from the grid or operating in parallel with it, needs battery energy storage to function reliably. In his capstone research, Sherman found this to be a consistent pattern across the case studies he analyzed: storage isn’t optional in these configurations; it’s structural.
Battery systems serve multiple roles in on-site data center power. They:
- Smooth the volatile load profiles generated by AI training workloads, which can swing rapidly between idle and peak draw.
- Provide near-instantaneous power injection during load ramps, protecting generators and maintaining power quality.
- Offer black-start capability and emergency backup.
- Participate in wholesale energy markets if the facility connects to the grid later.
“The battery isn’t merely an accessory in these installations,” Sherman says. “Without storage providing load smoothing and frequency support, an islanded power system can’t reliably serve the kind of workloads AI data centers demand. That makes battery technology selection a first-order strategic decision, not an afterthought.”
The Case for Looking Beyond Lithium-Ion
Most data center battery installations today rely on lithium-ion technology. It is commercially proven, energy-dense, and well understood by engineering teams. But drawing on a 2025 market report he completed during his MBA in Sustainable Innovation, Sherman found that the scale of AI-driven demand is exposing the limits of a single-chemistry approach.
Lithium-ion carries meaningful supply chain concentration risks. China refines approximately 59% of the world’s lithium, 68% of its nickel, and 73% of its cobalt. Price volatility, environmental concerns tied to extraction, and geopolitical tension around critical mineral supply chains add layers of risk that compound as deployment volumes grow. For a data center developer building multiple campuses that each require hundreds of megawatt-hours of storage, these are material planning variables.
In his research, Sherman focused on three alternative chemistries poised to reach commercial readiness at the right moment:
- Sodium-ion batteries, built from materials far more abundant than lithium, offer lower cost per kilowatt-hour and solid performance in cold climates.
- Iron-air batteries can discharge continuously for up to 100 hours, making them suitable for multi-day backup scenarios that lithium-ion cannot economically address.
- Flow batteries, such as iron-flow systems, provide modularity and 20-plus year lifespans without meaningful capacity loss, a compelling fit for facilities designed to run for decades.
“No single chemistry is going to cover the full range of what data center power systems need,” Sherman says. “Fast-response load smoothing, long-duration backup, and cost-effective bulk storage are three different problems. The companies and developers who treat them as one problem are going to overpay, underperform, or both.”
Where Energy Strategy Meets Community Impact
For Sherman, the AI data center power conversation carries implications that go well beyond the companies building the facilities. Federal policy has started to reflect this. In March 2026, Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI signed the White House’s Ratepayer Protection Pledge, committing to build, buy, or otherwise fund the power generation and grid infrastructure needed to run their data centers rather than shifting those costs onto households and businesses in surrounding communities.
The reasoning is straightforward: grid upgrades to serve data centers can drive up electricity costs for everyone else connected to the same system. “When a single facility requests 500 megawatts from the grid, the cost of the transmission upgrades required to deliver that power doesn’t stay inside the data center’s budget,” Sherman says. “It gets socialized. On-site generation paired with battery storage is one way to decouple that demand from the public grid, and the technology to do it reliably is available now.”
The question, in his view, is whether the battery storage industry can scale fast enough to match the pace of data center construction. With global AI investment accelerating and the U.S. and China engaged in a strategic competition over computing capacity, the demand for reliable, cost-effective, and rapidly deployable storage solutions isn’t likely to slow down.
This is where the questions he examined in his capstone research connect most directly to real deployment decisions being made now. “The companies that figure out how to deliver diversified storage solutions at the speed and scale this market requires are going to be in a very strong position,” he says. “And the communities where these decisions are being made deserve to be part of that conversation, not just downstream of it.”
