China’s First Wind-Powered AI Data Center Goes Live, Cuts 3.65 Million Tonnes of Carbon

China’s first AI data center powered entirely by wind energy is now operational in the Ningxia Hui Autonomous Region, according to a South China Morning Post report cited by Spanish technology media Xataka. The facility uses 100% renewable electricity, avoiding an estimated 3.65 million tonnes of carbon emissions per year based on the renewable plant’s annual generation of 4.3 billion kilowatt-hours.

The data center achieves a Power Usage Effectiveness (PUE) of 1.15, significantly better than the industry average of 1.56 over the past five years, as measured by the Uptime Institute’s Global Data Center Survey 2025. A PUE of 1.0 is considered perfect, meaning all energy used goes directly to computing equipment. The 1.15 figure indicates that 15% of total power is consumed by overhead such as cooling and lighting, a level that places the facility among the most efficient globally.

Key facts
| Metric | Value |
|——–|——-|
| Location | Ningxia Hui Autonomous Region, northwest China |
| Annual renewable energy generation | 4.3 billion kWh |
| PUE (Power Usage Effectiveness) | 1.15 |
| Annual carbon emissions avoided | 3.65 million tonnes |

A strategic shift: Eastern Data, Western Computing

The Ningxia installation is part of China’s national “Eastern Data, Western Computing” (EDWC) megaproject, which moves data processing from the densely populated eastern provinces to western regions rich in renewable energy resources. The strategy aims to reduce the carbon footprint of China’s fast-growing AI compute infrastructure while leveraging the west’s abundant wind and solar capacity.

The EDWC project is expected to help China meet a projected doubling of energy demand by 2030, according to forecasts cited in the original report. By relocating compute loads, the country can also reduce strain on coastal power grids and improve latency for non-time-sensitive AI workloads.

The energy reality of AI

The need for cleaner AI infrastructure is becoming urgent. Consulting firm Gartner estimates that AI data centers will consume 565 terawatt-hours of electricity in 2026, up from roughly 200 TWh in 2023. A recent study by Allianz found that emissions from AI data centers reached 286 megatonnes of CO2 in 2025, 57% higher than the International Energy Agency initially calculated. Electricity consumption accounted for 76% of that total footprint.

The Allianz report concludes that greening AI depends primarily on transforming the energy systems that power it, with wind power offering a particularly low water and carbon footprint. The Ningxia project provides a real-world demonstration of that approach at scale.

EU and US regulatory context

The European Union has already introduced energy transparency rules for data centers. The Energy Efficiency Directive (EU) 2023/1791 requires facilities with a connected load above 500 kW to publish their energy consumption. The Delegated Regulation (EU) 2024/1364 establishes a common evaluation framework for the more than 3,000 data centers in the EU.

Hongpeng Lei, head of the Mitigation Division at the UN Environment Programme (UNEP), described the shift toward sustainable data centers as “the path to a resilient and competitive digital infrastructure,” adding that efficiency should not be seen as a barrier to technological progress.

For the United States, which hosts the largest concentration of AI data centers, the Ningxia example raises questions about grid readiness. While the U.S. leads in renewable energy deployment, the scale of AI compute growth is straining regional grids. In Spain, a Microsoft project in Aragon is expected to consume more energy than the entire autonomous community. U.S. operators such as Google, Microsoft, and Amazon have committed to 24/7 carbon-free energy by 2030, but current grid constraints may make that target difficult without dedicated renewable microgrids similar to the Ningxia model.

What to watch next

The Ningxia facility is a single proof point, not a solution for the entire industry. Its PUE of 1.15, while excellent, still leaves room for improvement. The facility’s actual compute load and uptime performance have not been publicly disclosed. The EDWC project’s full impact will depend on how quickly China can build additional wind-powered data centers and whether the western grid can handle the transmission loads.

For U.S. readers, the key questions are whether American utilities can follow China’s lead in co-locating large-scale AI compute with dedicated renewable generation, and whether regulatory frameworks at the federal and state levels will accelerate or delay such investments.

Source: Xataka IA, citing South China Morning Post and Uptime Institute Global Data Center Survey 2025. Original article: https://www.xataka.com/energia/debatimos-cuanta-electricidad-necesita-ia-china-tiene-centro-datos-que-funciona-energia-limpia

Datos clave