Thirty Gigawatt-Hours of Iron

The largest battery ever announced, measured by energy capacity, will not contain a single lithium-ion cell.

Google and Xcel Energy disclosed an agreement on February 24 for 1,900 MW of new clean energy resources to power a data center in Pine Island, Minnesota. The package includes 1,400 MW of wind, 200 MW of solar, and a 300 MW iron-air battery system built by Form Energy. That battery will store 30 GWh of energy and discharge continuously for up to 100 hours.

The standard grid-scale lithium-ion battery discharges for two to four hours. Google ordered one that lasts four days.

The chemistry. Form Energy’s iron-air cells generate electricity by oxidizing iron pellets, a controlled rusting process that releases electrons. Charging reverses the reaction. The raw materials (iron, water, air) are among the most abundant and cheapest on Earth. Whether the company can manufacture cells at a cost that justifies the chemistry’s theoretical advantage over lithium iron phosphate remains unproven at commercial scale.

The 100-hour duration changes what a battery can do. A four-hour lithium-ion system shaves demand peaks and captures arbitrage spreads. It cannot keep a data center running through a three-day winter storm when wind and solar simultaneously underperform. A 100-hour system can. For facilities requiring 99.999% uptime, that is not an incremental improvement. It is a different product category.

The factory. Form Energy expects to begin deliveries to Xcel in 2028. No comparable iron-air manufacturing facility exists anywhere. The company must scale from pilot production to commercial output capable of supplying 30 GWh for a single customer, on a timeline measured in months rather than years. If Form delivers on time and on specification, it will have validated a battery chemistry that researchers have studied for decades but no one has manufactured at grid scale. If it does not, the 100-hour battery remains a laboratory result.

The ratepayer structure. Google will cover all grid infrastructure costs associated with the data center. The company also committed $50 million to Xcel’s Capacity*Connect program, which funds distributed storage assets on the broader Minnesota grid. Xcel plans to file an Electric Service Agreement with the state Public Utilities Commission for approval.

The deal is structured to prevent cost-shifting to existing Xcel customers. Whether that structure survives the regulatory process intact depends on the PUC filing, which has not yet been made public.

This structure landed on the same day as President Trump’s “Rate Payer Protection Pledge” at the State of the Union, which called on technology companies to “build their own power plants as part of their factory, so that no one’s prices will go up.” Google’s Pine Island agreement is, functionally, the template that pledge describes.

The customer. A technology company, not a utility, not a grid operator, not a government, placed the largest battery order in history. The order went to a startup that has never delivered a commercial-scale system. The chemistry is not lithium-ion. The duration is 25 times the industry standard.

This pattern extends beyond Google. The same week, Meta disclosed an agreement with AMD worth up to $100 billion for GPU chips that will require approximately 6 GW of new data center power capacity, equivalent to the electricity consumption of 4.5 million homes. ON.Energy and Shoals Technologies announced a partnership to deploy multi-gigawatt critical power systems into AI data centers, pairing medium-voltage battery UPS platforms with Shoals’ DC Recombiner technology, which consolidates inputs from solar, BESS, and other DC microgrid sources. Shoals is expanding into a 638,000-square-foot manufacturing campus in Portland, Tennessee to meet that demand.

Data centers are not adopting the battery industry’s existing product catalog. They are commissioning new ones. Google wants 100-hour iron-air for renewable firming. ON.Energy wants DC-coupled battery systems for critical power redundancy. Neither is a standard lithium-ion storage rack.

Two markets. Battery storage has been defined, until now, by a single dominant chemistry (lithium iron phosphate) in a narrow duration window (two to four hours). Data center demand is pulling in two directions simultaneously: much longer duration for energy security and much higher power density for critical infrastructure. Both require different manufacturing, different engineering, and different commercial structures than the grid-scale and commercial storage industry has built.

Peter Freed, who led energy strategy at Meta before founding advisory firm Near Horizon Group, argued in a Heatmap interview this week that data centers are “creating a new kind of battery monster,” splitting battery demand from its traditional association with renewable energy integration. The largest orders are not for energy shifting or peak shaving. They are for reliability at scale.

That bifurcation has consequences. Lithium-ion manufacturers competing for utility-scale and commercial contracts now face a future where the highest-volume, highest-value customers want products they do not make. Whether that drives lithium-ion incumbents to diversify into long-duration chemistries, or simply creates a parallel market with its own supply chain, depends on how quickly Form Energy and its competitors can prove that iron, water, and air can do what lithium cannot.

The largest battery order in history went to a company that does not use lithium, from a customer that does not operate power plants, for a chemistry that has never been manufactured at commercial scale. Every assumption embedded in that sentence has to hold for the bet to pay off. Google is wagering that it will.

Sources