Google Backs Multi-Tech Storage to Solve 24/7 Clean Power Gaps for AI Data Centers

Google, in partnership with renewable developer MN8 Energy and zinc battery manufacturer Eos Energy Enterprises, announced an integrated clean energy initiative in Kanawha County, West Virginia, dubbed Mammoth Solar. Sited on a reclaimed coal mine, the project pairs 86 MW of utility-scale solar generation with a hybrid storage portfolio consisting of 70 MW / 280 MWh of lithium-ion batteries and a 10 MW / 100 MWh 10-hour zinc-based long-duration energy storage (LDES) system. Google will purchase the project's energy, capacity, and clean attributes to support its regional data center infrastructure across the PJM Interconnection grid. Commercial deployment will be executed in stages, with solar operations starting in 2028, lithium-ion storage in 2029, and the zinc LDES system coming online in 2030. For DevOps leaders, platform architects, and sustainability officers, this development marks a shift in how hyperscalers address the heavy power appetite of modern workloads. As generative AI training and real-time inference drive data center energy density higher, cloud providers can no longer rely solely on intermittent solar contracts or standard four-hour battery discharge curves to guarantee clean operations. When solar output drops off, data centers routinely draw from fossil-heavy grid capacity. By incorporating 10-hour non-lithium storage, Google is bridging the overnight generation chasm, moving the industry closer to genuine hourly matching and providing enterprise clients with auditable Scope 3 emission reductions rather than annual market averages. This deployment fits into a broader, accelerating trend where cloud hyperscalers are investing directly into alternative, firm clean technologies to future-proof their operations. Over the past year, major operators including Google, Microsoft, and AWS have expanded their procurement portfolios beyond conventional wind and solar to encompass advanced nuclear power purchase agreements, enhanced geothermal systems, and multi-day thermal and chemical energy storage. With regional transmission operators like PJM facing capacity constraints and lengthy interconnection queues, hyperscalers are taking a hands-on role in co-designing hybrid grid assets that supply continuous dispatchable energy without overloading local utility networks. For enterprise technology practitioners, this multi-tier storage paradigm provides clear operational takeaways. First, architecture teams designing carbon-aware workload schedulers can anticipate more granular 24/7 carbon-free energy telemetry at the regional grid level, allowing batch jobs and heavy model training pipelines to run overnight without incurring high marginal grid emissions. Second, platform teams should expect hyperscalers to reflect the capital expenditure of advanced storage architectures in regional compute pricing tiers, making carbon-optimized workload placement a crucial FinOps competency. Practitioners should start baselining workload elasticity now, preparing to align asynchronous computing tasks with emerging multi-duration grid storage profiles.