A single new AI data center in Virginia can draw as much power as a mid-sized city, and the US grid is straining to keep up. A quietly growing answer isn't lithium-ion at all: it's water-based batteries, and engineers building them say they "have not ruled out" using them to turn entire data centers into on-site energy storage assets.
What is the Concept
Water-based batteries, more precisely aqueous flow batteries and water-based electrolyte systems, store energy using non-flammable, water-soluble chemistries such as iron, zinc, or saltwater electrolytes instead of the flammable organic solvents used in lithium-ion cells. Because the electrolyte is water-based, these systems are far less prone to thermal runaway, which matters enormously when you're installing megawatt-hours of batteries next to racks of GPUs running around the clock.
US companies including Form Energy (Somerville, Massachusetts), ESS Inc (Wilsonville, Oregon), and Eos Energy Enterprises (Edison, New Jersey) have spent the past several years commercializing iron-air and zinc-based aqueous batteries for long-duration, stationary storage. Unlike lithium-ion, which is optimized for short bursts of high-density power, these water-based systems are built to discharge steadily for 4 to 100+ hours, which is exactly the profile a data center needs to smooth out grid demand rather than just survive a short outage.
Why It Matters in United States (2025-2026 Context)
US electricity demand from data centers is projected by the Department of Energy and grid operators like PJM Interconnection to roughly double by 2030, driven almost entirely by AI training and inference workloads. Dominion Energy, which serves "Data Center Alley" in Loudoun County, Virginia, has already flagged multi-year interconnection queues because new data centers are requesting power faster than transmission upgrades can be built.
This is a direct cost and speed-to-market problem for founders and CTOs. A hyperscale or colocation data center that can't get grid capacity approved can lose 18 to 36 months before racks are even powered on. Water-based battery storage offers a workaround: instead of waiting years for new transmission lines, operators in Texas, Arizona, and Virginia can install on-site storage that charges during off-peak, low-cost hours and discharges during peak demand, reducing both grid strain and the demand charges that make up a large share of a data center's electricity bill.
How AI Is Changing This
The irony is that AI itself is both the cause of this energy crunch and the tool solving it. Machine learning models are now used to forecast a data center's exact power draw hour by hour, allowing operators to schedule battery charging around real-time electricity prices on wholesale markets like ERCOT in Texas or CAISO in California. This turns a data center from a passive power consumer into what utilities call a "flexible load" or even a grid asset that can sell stored energy back during demand spikes.
AI is also accelerating battery chemistry research itself. Materials science teams at US national labs and startups are using machine learning to simulate thousands of electrolyte combinations in days instead of years, which is part of why water-based chemistries have moved from lab prototypes to multi-hundred-megawatt commercial deployments faster than expected.
Real-World Examples
Form Energy has already signed agreements with US utilities including Georgia Power and Xcel Energy to deploy iron-air battery systems designed for 100-hour discharge, originally aimed at grid-scale renewable backup but increasingly discussed as a model for co-located industrial and data center loads. ESS Inc's iron flow batteries are being piloted at commercial and industrial sites in the Pacific Northwest specifically because they avoid the fire-suppression and insurance complications that come with large lithium-ion installations near occupied buildings.
Microsoft and Google have both publicly stated they are evaluating alternative long-duration storage technologies, including flow batteries, as part of their pledge to run data centers on 24/7 carbon-free energy rather than annual offset credits. Neither has committed to full-scale water-based battery deployment yet, but the phrase used by engineers close to these pilots, that they "have not ruled it out," reflects genuine technical validation, not just PR positioning.
Practical Insights / Actions
If you operate or advise a data center, colocation facility, or any energy-intensive US operation, start by getting an interconnection queue estimate from your regional utility before you finalize a site. If that queue exceeds 12 months, model the ROI of on-site aqueous battery storage against the cost of delayed revenue, not just against the battery's sticker price.
For SMEs and mid-market SaaS companies renting colocation or cloud capacity, ask your provider directly whether they use time-of-use energy arbitrage or on-site storage. Providers doing this can often pass through lower, more predictable pricing, which is a meaningful negotiating point when locking in multi-year hosting contracts.
Future Outlook
Expect water-based, long-duration battery storage to become a standard line item in US data center RFPs by 2027, particularly in states with constrained grids like Virginia, Texas, and Arizona. Regulatory tailwinds, including Inflation Reduction Act storage tax credits, will keep lowering the payback period, making this less a sustainability nice-to-have and more a straightforward cost and uptime strategy.
Companies like RP SoftTech help founders and operations teams model infrastructure and automation decisions like this against real cost and revenue impact, which becomes increasingly important as energy costs shift from a fixed line item to a variable one that smart automation can actively manage.
Conclusion
Water-based batteries won't replace lithium-ion everywhere, but for the specific problem of powering AI-hungry US data centers without waiting years for new transmission lines, they're a genuinely non-obvious and increasingly credible answer. Founders and infrastructure leaders who model this into their 2026 planning now will have a real cost and speed advantage over those who wait for the grid to catch up on its own.