Every major economy is confronting an identical challenge today. Artificial intelligence is gobbling up electricity at a rate that power grids were never built to support. In the US, capacity market prices in PJM, the nation’s largest grid operator, have surged more than ten times in just two years, with data-centre expansion pinpointed as a key culprit. In Europe, utility companies are rushing to modernise transmission networks quickly enough to match hyperscalers’ relentless appetite.
The International Energy Agency (IEA) forecasts that worldwide data-centre electricity demand could near 1,000 TWh by the decade’s close. Renewable energy supply is largely in place, but the ability to orchestrate it, through AI energy grid mapping at national scales, is what most nations still don’t have. But China has just built one.
A study published in Nature this week by researchers from Peking University and Alibaba Group’s DAMO Academy has produced something no country has accomplished before: a comprehensive, high-resolution, AI-generated inventory of a whole nation’s wind and solar installations, complete with the analytical framework needed to coordinate them as a single, unified system.
By training a deep-learning model on sub-metre satellite imagery, the team pinpointed China’s 319,972 solar photovoltaic sites and 91,609 wind turbines, processing 7.56 terabytes of imagery to achieve this.
AI energy grid mapping
Earlier research into solar-wind complementarity, the notion that the two sources can counterbalance each other’s fluctuations across time and geography, has largely depended on theoretical or modelled deployment scenarios. How complementarity actually plays out under real-world infrastructure conditions, and how it influences system-wide integration results, has until now been uncertain.
The researchers demonstrate that solar-wind complementarity cuts generation variability significantly, with effectiveness strengthening as the geographic range of pairing widens.
In practical terms, the more spread out the coordinated facilities are, the more dependable the balance becomes. A cloud blanketing solar farms in Gansu does not simultaneously block wind resources in Inner Mongolia, for example. The study’s findings highlight a structural inefficiency in how China currently runs its grid: coordination occurs at a provincial level rather than across the whole nation.
Shifting to a unified national approach, the researchers argue, would make it simpler to pair complementary energy sources, stabilise the grid, and reduce curtailment, the wasting of generated renewable power that has long been one of China’s most expensive clean-energy headaches.
Liu Yu, a professor at Peking University’s School of Earth and Space Sciences, described the inventory as enabling China to view its new-energy landscape from a “God’s-eye view,” a phrase that carries greater practical significance than it might initially seem. Grid operators cannot optimise what they are unable to see, until now.
China is in the grip of an AI-driven electricity demand boom that is stretching its grid to its limits. The rapid spread of data services and enormous computing facilities have driven the sector’s power consumption up 44% year-on-year in the first quarter of 2026, hitting 22.9 billion kilowatt-hours, according to the China Electricity Council.
That is a remarkable growth rate for a sector whose demand was already substantial. This has fast-tracked data-centre expansion in China’s northern and western provinces, where land is cheaper, wind and solar resources are more plentiful, and electricity prices are correspondingly lower. The provinces earmarked for new data centres are the same regions with the strongest solar-wind complementarity.
Behind the model
The technical feat behind this deserves recognition in its own right. DAMO’s deep-learning model was trained to detect solar photovoltaic facilities and wind turbines from sub-metre resolution satellite imagery, a task made difficult by the sheer variety of installation types, landscape conditions, and image quality.
The resulting dataset covers installations in 1,915 Chinese counties, spanning everything from rooftop panels on coastal cities to utility-scale wind farms on the Mongolian plateau. Processing 7.56 terabytes of imagery to create a nationally consistent, county-level inventory is a compelling demonstration of what large-scale geospatial AI can accomplish when applied to infrastructure challenges, and a blueprint that other nations could, in theory, replicate.
China’s clean energy sector produced an estimated 15.4 trillion yuan (US$2.26 trillion) in economic output last year, on a par with Brazil’s entire GDP, according to the Finland-based Centre for Research on Energy and Clean Air. Managing an asset base of that magnitude without a national-level visibility tool was always going to be a ceiling, a ceiling that has now been lifted.
The study’s dataset and code have been made publicly accessible via Zenodo.
(Photo by Luo Lei)
See also: Inside China’s push to apply AI in its energy system
Want to learn more about AI and big data from industry leaders? Check out AI and Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.
AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
