Retired EV Batteries Emerge as Grid Solution for AI Data Centers

The power crunch behind AI infrastructure

America's artificial intelligence data centers face a mounting electricity crisis. Hyperscale facilities running large language models and GPU clusters are straining power grids, triggering disputes over rates, and forcing utilities to reconsider how they serve some of the most energy-intensive infrastructure ever constructed. The hunt for affordable, reliable power has become as critical as the race to build the facilities themselves.

A promising answer may already exist in warehouses nationwide: retired electric vehicle batteries.

Battery packs removed during warranty replacements and routine maintenance typically retain roughly 80% of their original capacity. While no longer optimal for automotive applications, they remain well-suited for stationary energy storage that can support AI operations. The approach is straightforward—batteries charge during off-peak hours when electricity is cheaper, then discharge when AI workloads drive demand higher, creating a buffer between the facility and the grid.

Real-world deployment shows viability

In summer 2025, Redwood Materials deployed what it called North America's largest second-life battery storage system at its Sparks, Nevada campus. The 12-megawatt, 63-megawatt-hour microgrid, assembled from hundreds of repurposed EV packs, powers a modular data center operated by Crusoe Energy housing approximately 2,000 GPUs. The system has maintained 99.2% operational availability since launch. By early 2026, Crusoe announced plans to expand from four to 24 modular units, increasing power demand to roughly 20 megawatts.

The economics are shifting in favor of reuse. Bloomberg New Energy Finance found that new lithium-ion battery pack prices for stationary storage fell to approximately $70 per kilowatt-hour in 2025, down about 45% year-over-year. Repurposed packs sourced directly from automakers can cost significantly less.

Why it matters

Beyond immediate cost savings, battery reuse addresses a critical national security vulnerability. China controls roughly 90% of global bismuth supply—a metal increasingly vital for liquid-cooling systems in high-performance AI processors. The country also dominates approximately 85% of rare-earth processing capacity and receives roughly half of U.S. copper scrap for processing. Domestic battery repurposing reduces dependence on these concentrated supply chains while creating a scalable solution from an expanding waste stream. With global data-center battery market value projected to grow from $3.38 billion in 2025 to nearly $6 billion by 2035, and lithium-ion shipments for AI storage expected to surge from 12 gigawatt-hours in 2025 to 272 gigawatt-hours by 2030, the timing aligns with both supply and demand.

Challenges remain before widespread adoption

Data center operators invest heavily in uptime guarantees, making them cautious about any new power-chain component. Repurposed batteries vary in age, chemistry, and degradation rates, creating less predictable performance than new systems. Thermal runaway risks require rigorous testing, monitoring, and battery-management software. Certification standards and insurance requirements also present hurdles.

Linda Li, CFO of Re-Teck—a global IT asset disposition firm that receives retired packs from Tesla, Lucid, and BMW while serving Microsoft and other data center operators—acknowledges the company is still building the case studies needed to move from interest to contracts. "If you have to buy new batteries to do this, it will never be practical," Li said. "Data center operators are drawing energy from the grid 24/7, peak and off-peak. They know that if they have an alternative, like a storage unit, they can supplement peak-hour energy with off-peak energy."

The details were first reported by Forbes contributor Ken Silverstein.