Meta and Google Bet on Long-Duration Energy Storage to Power AI Data Centers
nashnova research
Meta and Google have each signed deals with energy startups to deploy battery systems that store power for tens to hundreds of hours, replacing diesel generators at AI data centers. This means → the explosive growth in AI compute is fast-tracking a clean-energy technology that had stalled for years.
What exactly did Google and Meta each sign up for?
Google is working with Italy's Energy Dome to deploy a CO₂-based storage system in Arizona that holds 10 hours of power. In plain terms = the system compresses and releases carbon dioxide to store and discharge electricity — no lithium batteries involved.
Meta signed with California's Noon Energy for a system that stores 100 hours of renewable energy. It works by splitting CO₂ into carbon compounds and oxygen to store energy, then reversing the reaction to generate power.
Amanda Yang, Meta's head of clean and renewable energy, put it plainly: "Four hours only solves a limited problem. Ultra-long-duration storage is what truly enables clean, reliable power."
How fast is demand growing?
BloombergNEF estimates that since early 2025, newly announced, approved, or financed long-duration storage projects in the U.S. total more than 150 GWh — over five times the country's entire planned capacity at the end of 2024.
This means → in roughly half a year, new additions alone exceeded five times the prior total pipeline. The growth is a step-change, not a gradual ramp.
Aric Saunders, EVP at Noon Energy, said AI data centers now account for roughly half of the company's planned deployments — up from nearly zero two years ago. In plain terms = long-duration storage went from "nobody's buying" to "AI is buying it all" in two years.
Why not just keep using lithium-ion batteries?
Lucia Tian, Google's director of advanced energy technology, noted that many long-duration storage technologies do not rely on critical minerals like graphite, sidestepping geopolitical supply risks. This reflects a widening anxiety among tech companies — supply-chain security concerns now extend from chips to battery raw materials.
Startups are also building factories on U.S. soil, giving tech firms both supply-chain resilience and political goodwill.
Cully Cavness, co-founder and president of AI data-center operator Crusoe, called long-duration storage "a critical tool for building AI infrastructure at unprecedented scale." Crusoe is working with California's Energy Vault on a gravity-based storage concept — in plain terms = lift a heavy weight to store energy, lower it to generate power.
How does the cost problem get solved?
Meta's Yang acknowledged that long-duration storage is currently more expensive, but said it will show "extremely strong cost competitiveness" once deployed at scale — though she did not disclose specific price comparisons.
For startups, AI data-center clients matter beyond the order itself. Massachusetts-based XL Batteries said a procurement contract from data-center developer Prometheus Hyperscale provided critical market validation.
This means → a contract from a tech giant doubles as a credit endorsement, helping startups secure follow-on financing. XL Batteries CEO Tom Sisto was direct: "The path to bankability is the challenge every emerging technology provider will face, and AI data centers are exactly the players that can help."
What is the real test for this sector?
Long-duration energy storage — technology that holds power for hours to days, unlike the common four-hour lithium-ion systems — had stalled for years. The core bottleneck: high costs and no large-scale buyer willing to commit.
The AI compute boom addresses both problems at once: tech giants bring large orders and capital endorsement, giving startups a shot at crossing the "valley of death" from lab to commercial scale.
This signals something deeper: AI is not only consuming energy — it is reshaping the energy industry's technology roadmap and capital structure. Whether long-duration storage can truly take off on the back of Big Tech's backing is the sector's defining proof point.
市场有风险，内容仅供研究参考，不构成投资建议。