From Beams to Bytes: The Case for Firm, Dispatchable Energy From Solar Thermal for Artificial Intelligence Data Centers

Abstract

Surging electricity demand from data centers has led to growth in interest and investment in new energy sources, both in the U.S. and internationally. In locations where demand growth may surpass grid capacity growth, data centers need firm power that is quick to deploy, demand flexibility that complies with utility rate structures and state regulations, and thermal load management that balances electricity availability with technology constraints. Concentrating solar power (CSP) exhibits potential to support all these needs. CSP has deployed several gigawatts globally, leading to significant cost declines and rising capacity factors, with deployment and operating experience leading to improvements in performance and reliability and development of standards. In this manuscript, preliminary analysis shows CSP is cost-comparable (under $100/MWh) with other energy assets and retail rates in key Southwestern U.S. markets at an estimated 80% capacity factor. Larger markets in the Eastern U.S. (e.g., Virginia) require hybridization with PV to reach the capacity factor requirements. Additionally, CSP (and related technologies) may be paired with long-duration thermal energy storage to provide flexibility and offer backup power during grid stress events. CSP as a front-of-the-meter asset can complement integration of other low-cost energy sources, reducing the cost of capacity expansion. Lastly, as a thermal technology, CSP can directly integrate with data center’s thermal loads, providing dispatchable cooling services to manage peak cooling loads. The CSP industry must overcome hurdles in the U.S. to meet this growing market such as growing a U.S. supply chain, unlocking further cost reductions through new media and heliostat designs, and integration of advancements in AI. For data center stakeholders, working with industry experts, diligent siting and planning, and new financing structures can support successful deployments.