Data and code for: Global utility-scale photovoltaics are structurally displaced from electricity demand centres

This record contains the data and code associated with the study “Global utility-scale photovoltaics are structurally displaced from electricity demand centres.” The study develops an integrated asset–supply–demand–mismatch–allocation framework to diagnose the spatial mismatch between utility-scale photovoltaic (PV) supply and electricity demand at the global scale. Using Sentinel-2 imagery and a lightweight deep neural network deployed through Google Earth Engine, the study maps utility-scale PV assets across land between 60°S and 60°N at 10 m resolution, and combines these observations with latitude-dependent installed-capacity density, long-term PV generation climatology, and a 1-km electricity-demand surface derived from nighttime lights and GDP elasticity. The results show that the mapped global utility-scale PV footprint reached approximately 17,523 km² in 2025, corresponding to an estimated 1,852 TWh of annual realized PV generation, compared with a global electricity demand of approximately 32.8 PWh. This implies a mean annual utility-scale PV penetration of only 5.73%, indicating that PV still plays a largely supplementary role in the present global electricity system. Provincial-scale coupling further reveals pervasive under-coverage: regions below interim PV-share thresholds account for the majority of global electricity demand and population. At the national scale, the study quantifies mismatch using an optimal-transport source–load separation metric. Across the 120-country sample, the minimum average connection distance required to link PV supply to electricity demand centres ranges from 2.87 km to 1,980 km, with a median of 194.62 km, indicating that the current PV transition is increasingly constrained not only by deployment, but by the system capacity required to connect spatially displaced supply to load centres. Under IEA-aligned PV-share scenarios, meeting higher targets requires both larger PV supply volumes and greater reliance on domestic redistribution, selective cross-border corridors, and developable resource potential beyond existing production centres. Together, these results suggest that the limiting factor for scaling PV is no longer generation potential alone, but the capacity for spatial system integration, including transmission, grid access and flexibility. This Zenodo record includes the main materials needed to reproduce the core analytical workflow and reported outputs, including: a global vector dataset of mapped utility-scale PV distribution for 2025; Python code for local deep neural network training and export of model parameters for Earth Engine deployment; Google Earth Engine code for partitioned global inference of PV probability rasters; Python code for local post-processing of partitioned PV probability rasters; scenario output tables for candidate transmission directions under 8.6% and 15.7% PV-share targets. These materials are intended to support transparency, reproducibility and reuse of the spatial workflow developed in the study.