US adds 9.7 GWh storage

The U.S. installed 9.7 gigawatt-hours of energy storage in the first quarter of 2026, the strongest first quarter on record, according to Solar Energy Industries Association and Benchmark Mineral Intelligence data released this week. SEIA said the total was up 32% from the same period a year earlier and brought cumulative installed U.S. storage to 175 GWh. (seia.org) That number matters because the market around artificial intelligence is shifting from chips alone to the physical inputs needed to run data centers. (seia.org) Recent company announcements and investor coverage show power supply, land and grid access moving to the center of the build-out for AI facilities. ### Why is battery storage showing up in an AI story? SEIA said demand for battery energy storage systems has been supported by lower system costs and rising needs for reliability and flexible power. In practice, that lines up with what AI developers and cloud infrastructure companies are now confronting: new computing clusters need electricity that is available on time, not just hardware that can be ordered. (seia.org) CNBC reported on May 21 that a basket of AI-linked energy and infrastructure companies had outperformed Nvidia over the last year, reflecting investor interest in the power side of the AI expansion. The report framed the trade around companies exposed to generation, transmission and physical build-out rather than only semiconductors. (cnbc.com) ### What is the bottleneck for new AI data centers? Nebius Group said this week it signed a 10-year agreement with Bloom Energy to deploy fuel-cell power across its U.S. AI data centers. Yahoo Finance reported the deal carries a lifetime value of about $2.6 billion, while DatacenterDynamics said Nebius chose Bloom for faster time-to-power as it expands AI infrastructure. (seia.org) That arrangement points to the current constraint: developers are looking for power that can bypass or reduce delays tied to grid interconnection. Behind-the-meter generation, batteries and on-site power systems have gained attention because data-center projects can be held up by transmission limits, utility queues and local permitting. (cnbc.com) That is an inference from the structure of the Nebius-Bloom deal and related reporting on AI infrastructure, rather than a direct quote from one source. ### How are Nvidia and IREN part of the same trend? Nvidia and IREN announced earlier this month a partnership to accelerate deployment of up to 5 gigawatts of AI infrastructure. (finance.yahoo.com) Nvidia said the arrangement combines its AI factory architecture with IREN’s capabilities in power, land, data centers, GPU deployment and infrastructure operations. Reuters and CNBC reported that the agreement also gives Nvidia the right to invest up to $2.1 billion in IREN through a five-year share purchase right. The structure underscored that access to powered sites and deployable infrastructure is now valuable enough to sit alongside chip supply in AI transactions. (finance.yahoo.com) ### What does the 9.7 GWh figure say about the broader build-out? SEIA said the United States has now installed enough storage capacity to power 5.3 million homes for a year on a single charge, using its stated comparison. The trade group also said installed storage has risen 177% since the end of 2023. (nvidianews.nvidia.com) Those figures do not mean batteries alone can solve AI’s power needs. They do show that storage is scaling quickly at the same time AI developers, utilities and infrastructure operators are competing for generation, transmission access and construction-ready land. That is the backdrop for why energy assets are increasingly appearing in AI financing and partnership announcements. (money.usnews.com) ### What should readers watch next? SEIA and Benchmark’s next market updates will show whether the first-quarter pace continues through the rest of 2026. Company announcements are likely to provide the clearest near-term signals, especially deals involving on-site generation, storage, utility service agreements and data-center campuses with secured power. (seia.org) For now, the clearest fact is that 9.7 GWh was not just a storage record. It arrived as AI infrastructure builders, from Nebius to IREN, were signing deals that put electricity supply and site readiness at the center of expansion plans. (seia.org)