Google and Xcel Energy just announced the largest grid battery ever proposed: 30 gigawatt-hours of storage built to run for 100 hours straight by rusting iron on command. It is a bet that patience, not speed, is what lets solar and wind power an AI data centre around the clock.
It is a cold, still night in rural Minnesota. The wind has barely turned the turbines for three days straight, the sun set behind a cloud bank hours ago, and a grid operator somewhere is eyeing the gas plants on standby. Ordinarily this is when things get tense.
Not here, not soon. Under a field near Pine Island, Minnesota, engineers plan to bury a battery that stores electricity by deliberately rusting iron, then unrusting it to get the power back. It will not care that it is night three of a windless spell. It is designed to keep discharging for more than four days straight.
That is the idea behind the biggest grid battery ever announced: a bet by Google and Xcel Energy that the way to run an always-on data centre on wind and solar is not a faster battery. It is a patient one.
Here is what happened
The largest ever announced. On February 24, 2026, Google and Xcel Energy unveiled plans for a 300-megawatt, 30-gigawatt-hour battery in Pine Island, Minnesota, about 70 miles southeast of Minneapolis-St. Paul. Xcel calls it the largest battery project by energy capacity ever announced anywhere in the world. It uses iron-air technology from Form Energy, a Massachusetts startup, and is built to discharge at full power for up to 100 hours, more than four days without a recharge.
Part of a bigger deal. Google is paying Xcel to add 1,400 megawatts of wind and 200 megawatts of solar to feed a new data centre that Xcel says will support everyday services like Search, Gmail, YouTube and Maps. Google covers the new grid costs and is putting 50 million dollars into a separate program to ease local congestion, and Xcel says existing customers will not pay more because of the deal.
Built on a former steel mill. The batteries will be made at Form Energy's 550,000 square foot factory in Weirton, West Virginia, on a former steel mill site that is on track to reach 500 megawatts of yearly production by 2028. Form expects to ship the first Pine Island modules that same year, with installation phased in through 2031.
Already proven through a Minnesota winter. This is not Form's first Minnesota project. A smaller 1.5-megawatt, 150-megawatt-hour pilot with the cooperative Great River Energy broke ground in 2024, and by early 2026 Form said it had run a fielded system there for more than 100 hours straight through a sub-zero winter, its first proof outside the lab that the technology holds up.
How it works
Rust, run in reverse. Iron-air storage runs on one of the oldest reactions there is: rust. Charging means running electricity through iron pellets sitting in water, which strips the oxygen off them and leaves bare metallic iron behind, essentially undoing rust. Discharging means letting the iron breathe: it reacts with oxygen pulled from the air, turns back into rust, and that reaction releases electrons as usable electricity. Charge it, the iron stays bare metal. Run it, the iron rusts.
Cheap by design. Iron is one of the cheapest, most abundant metals on Earth, worlds away from the lithium, nickel and cobalt that make conventional batteries expensive to build at scale. Form Energy has said its goal is capacity costs a small fraction of lithium-ion's, because the whole design point is cheap hours of storage, even if the power delivered in any one hour is modest.
Duration, not punch. A lithium-ion battery unloads a lot of power fast and recharges quickly, which is why it is well suited to the four-hour bursts that smooth out an ordinary day. Iron-air charges and discharges slowly and is poor at that job. What it is built for is sitting there for days, discharging steadily through the kind of dark, calm, cloudy stretch that would otherwise force a grid to lean on gas. Duration, not power, is the entire point.
Why it matters
AI is the reason, right now. Data centres used to be a quiet, steady load. That is changing: the International Energy Agency projects data centre electricity use worldwide could roughly double this decade as AI scales up, and unlike a factory, a data centre wants power every hour of every day. Wind and solar are cheap but intermittent, so utilities increasingly fill the gap with an old, dirty standby, the peaker plant, a gas unit built for short bursts that is being pressed back into service to meet AI-driven demand. Xcel says the Form battery is meant to do a peaker's job without the emissions or the turbine supply delays.
Lithium-ion cannot close the gap alone. It excels at short, sharp bursts, but its economics worsen the longer it holds power, and past roughly four to six hours you are paying for expensive cells to sit mostly idle. That four-hour wall is exactly what long-duration storage, any technology built to discharge for eight hours or more, is trying to break through.
A crowded, unsettled race. Eos builds cheaper, non-flammable zinc batteries for 3-to-12-hour jobs. ESS makes iron flow batteries and has a Google-backed pilot in Arizona. Antora stores power as heat in glowing carbon blocks hot enough to melt steel. Hydrostor compresses air into underground caverns for an 8-hour California plant. China is scaling sodium-ion, led by CATL, as a cheaper lithium alternative. And pumped hydro, water pushed uphill and released through turbines, still quietly holds more than 90 percent of all long-duration storage on Earth, even though new dam sites are scarce. Iron-air's pitch is that it can go almost anywhere a dam cannot.
The honest catch
Strip away the excitement about rust saving the grid, and three hard limits remain.
Efficiency is weak. Iron-air batteries return only about 40 to 50 percent of the electricity put into them, versus 85 to 90 percent for lithium-ion. Put in 10 megawatt-hours, get back only 4 to 5. Battery consultant Jim McDowall has noted that at a real 40 percent efficiency, fully charging a 100-hour battery can take 250 hours, limiting it to roughly 25 full cycles a year.
It is slow by design. Iron-air cannot ramp up or down quickly, so it will not replace the lithium-ion systems that catch sudden spikes in demand. Grids heading this way will need both technologies, not one instead of the other.
Mostly promises, not power. Form's flagship Minnesota pilot only proved itself in 2026, Pine Island will not ship its first modules until 2028, and cost claims from Form and rivals like Hydrostor and Antora remain largely unverified at full commercial scale.
EDITOR'S TAKE
The number that matters here is not 30 gigawatt-hours, it is 100 hours. Almost every battery on the grid today is built for a four-hour evening; iron-air is a bet on the rare, brutal, multi-day lull that is the real thing standing between us and a grid run on sun and wind. The 40 percent round-trip efficiency looks damning until you remember the fuel is free solar and wind you would otherwise spill, and iron is close to free. So the question is not whether rust is elegant, it plainly is not, but whether Form's costs hold once the subsidies and the headlines thin out. Watch the 2028 first modules and the price per kilowatt-hour, not the ribbon-cutting. If the economics land, rust quietly becomes infrastructure. If they do not, this is a very large science project with Google's name on it.
Quick questions
What is an iron-air battery, in plain terms?
It is a battery that stores electricity by controlling how iron rusts. Charging strips oxygen from iron using electricity, leaving bare metal behind; discharging lets that iron react with oxygen from the air and rust, releasing electricity as it does. Form Energy is furthest along in commercializing it, with a factory in Weirton, West Virginia, and pilot and utility-scale projects underway in Minnesota. It is cheap to build because iron is abundant, but it charges and discharges slowly, so it is built for multi-day duration, not quick power.
How is this different from a normal lithium-ion battery?
Lithium-ion batteries are built for speed: they charge and discharge quickly and return about 85 to 90 percent of the electricity put in, but they get expensive fast once asked to run for more than about four to six hours. Iron-air batteries do the opposite. They are slow, only return about 40 to 50 percent of stored electricity, and are poor at sudden bursts, but they can discharge steadily for up to 100 hours because their raw materials are so cheap. Most grids will likely end up using both, lithium-ion for daily swings and iron-air or similar tech for multi-day gaps.
When will the Google and Xcel Minnesota battery actually be running?
Not soon. Form Energy expects to ship the first battery modules for the Pine Island, Minnesota project by the end of 2028, with the full 300-megawatt, 30-gigawatt-hour installation phased in through 2031. A smaller, earlier Form pilot with Great River Energy in Minnesota has already run for more than 100 hours straight, which is the main evidence so far that the technology performs outside a lab. Until Pine Island is built and running, "world's largest" describes an announced plan, not power actually flowing to the grid.
Sources
World's largest grid battery part of Google-Xcel Energy agreement: Trade-press report with the deal terms, MW and GWh figures, and Form CEO quotes.
Google Minnesota data centre deal includes 30GWh multi-day iron-air batteries from Form Energy: The wind, solar and battery package behind the new data centre.
Gigantic Form Energy battery to power Google data center in Minnesota: Reporting on the Pine Island location, timeline and ratepayer protections.
Google bets big on 30GWh of Form Energy's iron-air storage despite efficiency trade-offs: Deep dive on round-trip efficiency, with Xcel and consultant quotes.
Great River Energy and Form Energy break ground on first-of-its-kind multi-day energy storage project: Form's own account of its first Minnesota pilot.
Long-duration energy storage deployments rose 49 percent in 2025: WoodMac: Market-wide context on the broader long-duration storage race.
Frontier Signal explains frontier technology in plain English. Company and agency figures should be independently verified. This is general information, not investment or professional advice.
