Google and Xcel Energy announced on February 24, 2026 a new data center in Pine Island, Minnesota powered by a 300 MW (30 GWh) Form Energy iron-air battery—the world’s largest by energy capacity. The 100-hour storage can power operations for days, not the typical 4 hours of lithium-ion. With AI workloads straining grids nationwide, Google is betting multi-day backup at under $20/kWh beats nuclear reactors. Installation starts in 2028.
100 Hours Changes the Math
Form Energy’s iron-air battery stores energy for 100 hours—enough to cover multi-day gaps when wind stops and clouds block solar. Lithium-ion taps out at 4 hours, handling peak demand but requiring fossil backup for extended weather events. Moreover, Google’s 30 GWh system costs ~$20/kWh versus $130-150/kWh for lithium-ion. That 7x advantage makes week-long renewable coverage viable.
The technology uses reversible rusting: iron reacts with oxygen and water to form rust (iron hydroxide), releasing electrons. Charging reverses it. However, round-trip efficiency is 40-50%—you lose half the energy—but that beats lithium-ion’s economics for long-duration storage and hydrogen’s worse efficiency.
Furthermore, Form Energy’s Weirton, West Virginia plant started shipping in 2024. Google’s Minnesota deployment is the first grid-scale installation. The battery modules use iron, water, and air in shipping container-sized enclosures. No exotic metals, no supply chain constraints, non-flammable.
The AI Energy Crisis Hits in 2026
Data center electricity consumption hits 1,050 TWh globally in 2026, up from 415 TWh in 2024. Specifically, AI servers drive 30% annual growth versus 9% for conventional workloads. The grid wasn’t built for this—70% of U.S. infrastructure approaches end-of-life, and interconnection queues delay projects for years.
Consequently, PJM Interconnection, serving 65 million across 13 states, projects a 6 GW shortfall by 2027. Northern Virginia data center market is closed to new large-scale builds. Occupancy rates exceed 95% in late 2026.
Meanwhile, Google’s 1.9 GW package—1,400 MW wind, 200 MW solar, 300 MW battery—adds new capacity rather than competing for scarce supply. Notably, Google pays 100% of costs with zero ratepayer burden. Pending Minnesota Public Utilities Commission approval.
Why Iron-Air Instead of Nuclear?
Other hyperscalers bet differently. For instance, Microsoft signed a 20-year PPA to restart Three Mile Island nuclear reactor and tested hydrogen fuel cells for 48-hour datacenter runtime. Meta built one of the largest lithium-ion systems for Mesa datacenter. AWS hit 100% renewable matching in 2024 through wind, solar, and conventional batteries.
Google chose iron-air because nuclear requires 5-10 years for approvals and construction. Additionally, hydrogen offers unlimited duration but 60-75% round-trip losses versus iron-air’s 50-60%. Lithium-ion can’t economically cover multi-day gaps.
Geography matters too. Minnesota has strong wind resources. Pumped hydro needs elevation and water. Compressed air needs underground caverns. In contrast, iron-air works anywhere—add shipping containers for capacity, no site-specific engineering.
Form Energy’s Big Bet
Form Energy was founded in 2017 by Mateo Jaramillo (former Tesla battery development head) and Yet-Ming Chiang (MIT professor, A123 co-founder). The startup raised $240 million and invested $760 million in its Weirton plant. Therefore, Google’s Minnesota project is the first grid-scale deployment—unproven at commercial scale.
The risk is real. If the system underperforms, Google needs fossil backup or nuclear PPA. The 2028 timeline allows little delay. Furthermore, 40-50% efficiency requires overbuilt renewable generation.
If it works, economics reshape the industry. Importantly, a March 2025 Nature Energy study found storage must hit $20/kWh to cut carbon-free system costs by 10%. Form Energy claims to be there. AWS, Microsoft, and Meta are watching. Success means iron-air announcements from competitors by 2029.
The 2030 Test
Google is betting over $1 billion that reversible rusting beats nuclear or hydrogen for datacenter sustainability. The Minnesota site will support Workspace, Search, YouTube, and Maps—core services requiring high uptime. By 2030, operational data proves whether 100-hour batteries deliver on economics and reliability.
The broader stakes extend beyond Google. Iron-air at $20/kWh unlocks higher renewable penetration by eliminating multi-day intermittency. Utilities retire gas peakers, island grids cut diesel, industrial facilities decarbonize backup. All hinges on 2028 success.
No one else is taking this bet. Google is either visionary or overconfident. We’ll know in two years.
