The potential of agrivoltaic systems to meet energy demand of data centers in the United States

Artificial intelligence (AI) use has become so prevalent that AI-based energy use is straining the electrical system and impacting large-scale electrical grid planning. Calls for sustainable AI come at a time when solar photovoltaic (PV) provides the lowest cost electricity in history, and is also experiencing incredible growth. Both of these growth sectors, however, have had social concerns raised because of land use conflict with farming. Fortunately, agrivoltaics provides a solution to this land use conflict for PV. This study provides the first analysis of agrivoltaic systems’ potential to also support data center electricity loads across different U.S. states to help mitigate this issue. Using state-level data center energy consumption and modeled agrivoltaic generation potential, the analysis explores how much of the digital sector’s demand could realistically be met with agrivoltaics as well as how much farmland would need PV investments to cover AI loads within the states that have the largest data centers. The results show that for 26 states with substantial data center activity, single axis tracker (SAT) agrivoltaics could produce 114–17,566 TWh, whereas vertical agrivoltaics produced 30–4148 TWh. Vertical installations required only between 0.003–2.124% of farmland across the targeted states, whereas SAT installations required 0.001–0.548%. It is clear that agrivoltaics provides a technically viable method of meeting the energy demands of AI growth in the U.S. while increasing food production for a wide range of crops already shown to benefit from agrivoltaic microclimates.