Redwood Materials pivots from battery recycling to data-center storage

Tech’s newest “AI boom” story is no longer about model weights or benchmarks. It’s about amperage.

TechCrunch reports that Redwood Materials—founded by former Tesla CTO JB Straubel—has turned energy storage into its fastest-growing unit, explicitly because AI data centers are colliding with grid reality. A year ago, Redwood didn’t have an energy-storage business; now it is building systems for data centers using “second-life” EV batteries that aren’t yet ready for recycling. The company says its San Francisco R&D footprint has expanded four-fold to a 55,000-square-foot facility with nearly 100 staff, funded in part by a $425 million Series E that included Google and Nvidia.

The numbers are the point. Redwood’s first disclosed deployment for Crusoe—a modular data center builder—delivers 12 MW with 63 MWh of capacity. That’s not a cute backup battery; it’s infrastructure. And Redwood says the pipeline now includes projects in the “hundreds of megawatt hours” and even “multiple gigawatt hours.” Translation: the AI race is becoming an energy-storage arms race, because interconnection queues are now measured in years. Redwood’s business development lead told TechCrunch that data center developers are being told it will take “five-plus years” to connect to the grid.

Meanwhile, Google is selling the opposite narrative: that AI is the cure for the malware problem. In a separate TechCrunch piece, Google claims its AI-driven defenses reduced policy-violating apps published on Google Play in 2025, blocking 1.75 million apps (down from 2.36 million in 2024). It says it banned over 80,000 developer accounts and now runs more than 10,000 safety checks per app. Google Play Protect, it adds, identified 27 million new malicious apps—up from 13 million the year before—suggesting attackers are increasingly bypassing the Play Store.

Put the stories together and you get the real industrial logic: the same companies pitching AI as “safety” are financing the physical buildout that makes AI a permanent baseload customer. Data centers don’t behave like office loads. Their power density is high, their duty cycles are relentless, and their capex is front-loaded. When the grid says “come back in five years,” the market’s answer is batteries, gas, and anything else that can be deployed without permission slips.

Redwood’s pitch—repurposing EV packs—sounds circular-economy virtuous. It may also be brutally pragmatic: second-life batteries are a way to buy time while utilities expand substations and transmission. But the governance question is unavoidable. Who bears the externalities when hyperscalers and their suppliers effectively build parallel power systems? What happens when “AI safety” becomes a branding layer for an energy sprint that locks in new industrial footprints before communities can even argue about zoning?

The takeaway is simple: the bottleneck is not innovation, it’s centralized infrastructure and the permitting state. The market is routing around it—again—this time with containerized compute and gigawatt-hour batteries.