Major economies worldwide are currently grappling with an unprecedented challenge: artificial intelligence is consuming electricity at a rate that existing grids were never designed to accommodate. This surging demand has led to significant increases in capacity market prices, such as the more than tenfold rise in PJM, the largest grid operator in the US, largely driven by data-center growth. European utilities are also struggling to rapidly upgrade transmission infrastructure to keep pace with hyperscalers' demands. The International Energy Agency (IEA) projects that global data-center electricity consumption could approach 1,000 TWh by the end of this decade. While renewable energy sources are increasingly available, the critical missing component for most countries has been the ability to coordinate these sources effectively at a national scale through AI energy grid mapping. China, however, has recently achieved a significant breakthrough in this area.

A groundbreaking study published in Nature by researchers from Peking University and Alibaba Group’s DAMO Academy has unveiled what no other country has managed before: a complete, high-resolution, AI-generated inventory of China's entire wind and solar infrastructure, coupled with an analytical framework to coordinate it as a unified system. Utilizing a deep-learning model trained on sub-metre satellite imagery, the team meticulously identified 319,972 solar photovoltaic facilities and 91,609 wind turbines across China, a process that involved analyzing 7.56 terabytes of imagery data.

This innovative approach to AI energy grid mapping addresses a crucial gap in prior research on solar-wind complementarity. While the concept that these two sources can offset each other's variability in time and geography has been explored, its manifestation under real-world infrastructure and its impact on system-level integration outcomes remained unclear. The researchers demonstrated that solar-wind complementarity substantially reduces generation variability, with its effectiveness increasing as the geographic scope of pairing expands. This means that facilities located further apart can more reliably achieve balance; for instance, cloud cover over solar farms in Gansu would not affect wind corridors in Inner Mongolia.

The study's findings also brought to light a structural inefficiency within China's current grid management, which operates at a provincial rather than national level. The researchers argue that transitioning to a unified national scale would significantly facilitate the pairing of complementary energy sources, thereby stabilizing the grid and mitigating curtailment – the costly waste of generated renewable power that has long plagued China’s clean-energy sector. Liu Yu, a professor at Peking University’s School of Earth and Space Sciences, likened the new inventory to offering China a “God’s-eye view” of its new-energy landscape, enabling grid operators to optimize resources they were previously unaware of.

China itself is in the midst of an AI-driven surge in electricity demand that is straining its grid. The rapid proliferation of data services and massive computing facilities led to a 44% year-on-year increase in the sector's power consumption during the first quarter of 2026, reaching 22.9 billion kilowatt-hours, according to the China Electricity Council. This extraordinary growth rate has accelerated data-center expansion into China's northern and western provinces, regions known for cheaper land, abundant wind and solar resources, and consequently lower electricity prices. Coincidentally, these are also the provinces identified as having the highest solar-wind complementarity, making the national coordination framework even more critical.

The technical achievement behind this project is remarkable. DAMO’s deep-learning model was meticulously trained to identify diverse solar photovoltaic facilities and wind turbines from sub-metre resolution satellite imagery, navigating complexities arising from varied installation types, terrain conditions, and image quality. The resulting dataset encompasses installations across 1,915 Chinese counties, from rooftop panels in coastal cities to vast utility-scale wind farms on the Mongolian plateau. This effort, processing 7.56 terabytes of imagery to create a nationally consistent, county-level inventory, stands as a testament to the power of large-scale geospatial AI in addressing infrastructure challenges and serves as a replicable template for other nations.

China's clean energy sector generated an estimated 15.4 trillion yuan (US$2.26 trillion) in economic output last year, an amount equivalent to Brazil’s entire GDP. Managing an asset base of this magnitude without a national-level visibility tool has always been a limiting factor, a constraint that has now been overcome. The study's dataset and code have been made publicly available via Zenodo.