Modular micro-data centers in cooler climates: Geography, strategy and California's role

Dr. Choi, an Associate Professor at Michigan Technological University, and Roberto "Bobby" Escobar, general counsel and an environmental advisor, authored an article detailing the evolving geographic strategy for data center deployment, driven by the explosive growth of AI and cloud services. The piece emphasizes the shift towards modular micro-data centers (MMDCs) that can be strategically placed in cooler climates to mitigate the significant energy and water demands of modern data infrastructure. Cooling, particularly for high-density AI workloads, is identified as a defining operational challenge, with traditional air-cooling systems consuming up to 40% of energy. The authors note that reliance on free-air cooling makes consistently cool climates highly desirable, pointing to Nordic countries and U.S. regions like the Pacific Northwest and northern Midwest as optimal locations due to renewable energy sources, cool temperatures, and lower land costs. Water scarcity is also a growing concern, with Phoenix-area data centers consuming hundreds of millions of gallons annually and a UC Riverside study projecting substantial increases in peak water demand for U.S. data centers by 2030, potentially necessitating billions in infrastructure costs. California presents a complex strategic tension for data center development. Despite high electricity prices, grid decarbonization mandates, and extreme summer temperatures in inland regions like the Central Valley and Inland Empire, the state remains an indispensable demand center for latency-sensitive workloads. Governor Gavin Newsom's October 2025 veto of legislation mandating water usage disclosure for data centers underscores the contentious nature of the issue. The article proposes a three-tier distributed architecture: heavy AI training in cool, energy-rich regions; smaller, liquid-cooled California nodes for latency-sensitive applications; and hyper-edge deployments for time-critical processing. This model, supported by modular infrastructure, allows for dynamic workload placement based on energy costs, cooling efficiency, and latency requirements, thereby optimizing data center operations and addressing the challenges posed by climate and resource constraints.