Meta and Google are pouring billions into long-duration batteries to keep AI running

Meta and Google are placing major bets on long-duration energy storage systems, the kind that can keep power flowing for days instead of hours, as they scramble to find enough clean electricity to feed their ballooning data center empires.

Standard lithium-ion batteries typically discharge for around four hours. The new wave of investments targets systems capable of 10 to over 100 hours of continuous discharge, a fundamentally different class of energy infrastructure.

The deals on the table

Google’s February 2026 agreement with Form Energy stands out as the marquee transaction. The deal covers a 300 MW / 30 GWh iron-air battery system slated for a Minnesota data center, reportedly the world’s largest by energy capacity. The project carries an estimated price tag of around $1 billion and promises 100-hour discharge capability.

Iron-air batteries work by essentially rusting iron to store energy and then reversing the process to release it. They’re far cheaper per kilowatt-hour than lithium-ion at long durations, though they’re bulkier and slower to respond.

In September 2026, Google partnered with MN8 Energy on what’s being called the Mammoth Solar project. That initiative combines 86 MW of solar generation with 10 MW / 100 MWh of zinc-based long-duration storage, with phased deployment dates stretching from 2028 to 2030. The project sits in West Virginia, with Eos Energy contributing to the storage component.

Across the Atlantic, Google has a 23 MW / 200 MWh CO2 battery project with Energy Dome planned for 2028 in Ireland. CO2 batteries store energy by compressing carbon dioxide into a liquid and then releasing it through a turbine.

Meta’s April 2026 agreement with Noon Energy reserves up to 1 GW / 100 GWh in ultra-long-duration storage. The arrangement starts with a more modest 25 MW / 2.5 GWh pilot project targeted for 2028.

Why conventional storage falls short

Lithium-ion batteries dominate the current grid storage market because they’re proven, relatively cheap for short durations, and manufactured at enormous scale. But their economics deteriorate rapidly as you extend discharge times. Doubling the duration roughly doubles the cost, since you need twice as many battery cells.

Technologies like iron-air, zinc-based, and CO2 batteries decouple power capacity from energy capacity in ways lithium-ion cannot. The storage medium itself, whether it’s iron pellets, zinc electrodes, or compressed CO2, is cheap. You scale energy by adding more of the cheap stuff rather than more of the expensive electrochemical cells.

The grid reliability angle

Google’s Mammoth Solar project, with its integrated long-duration storage, effectively replaces retiring dispatchable coal generation with dispatchable renewable generation in a region transitioning away from coal.

Most of these projects won’t be fully operational until 2028 or later. Form Energy has yet to operate its iron-air system at anything close to the 30 GWh scale Google is ordering.