Spatial Dynamics and Determinants of Data Center Distribution in China: Evidence from Geodetector Modeling

Data centers serve as the backbone of modern digital life. Every message stored, every stream watched, every algorithm trained—it all traces back to these vast facilities scattered across the world. However, their physical operations present significant challenges, including substantial energy consumption and growing carbon emissions, which have become a global concern. What was once a quiet technical concern has turned into a genuine environmental challenge and recent estimates suggest that data centers now consume roughly two percent of the world’s total electricity, and that share could climb to nearly eight percent within the next decade. The numbers alone explain the growing unease among policymakers. In China, the issue feels especially urgent because the country’s digital infrastructure has expanded rapidly, and by 2025 its data centers are expected to draw more than four percent of national electricity—enough to emit close to a billion tons of carbon. Most of these facilities are clustered in the high-energy-demand eastern provinces (e.g., Beijing, Shanghai, Guangzhou). Meanwhile, the western region possesses abundant wind and solar resources, representing significant potential for sustainable development. To optimize the national layout and promote coordinated regional development, the Chinese government launched the Eastern Data Western Computing project. This national strategy aims to facilitate the westward movement of computing demand, effectively aligning the growth of the digital economy with environmental sustainability. Despite such policy advances, research on the spatial distribution of data centers in China remains fragmented. Most existing studies rely on visual mapping and qualitative analyses, often extrapolated from developed economies that differ substantially in economic geography, energy policy, and environmental constraints. Quantitative and region-specific assessments that integrate spatial clustering with causal modeling are rare. Moreover, prior work has tended to overlook emergent sustainability variables—such as renewable energy access, carbon intensity, and temperature conditions—now critical to data center siting under national carbon neutrality targets. To this account, a new research paper published in Energy and Buildings and conducted by Professor Donglin Chen and Ph.D. Candidate Lei Wang from Wuhan University of Technology, Lecturer Mengdi Yao from Wuhan University of Science and Technology, alongside Guolong She from The Hong Kong Polytechnic University, the researchers applied two complementary analytical models: a spatial-statistical model to map and quantify clustering of data centers using indices such as nearest-neighbor and Moran’s I, and a influencing factor analysis model employing Geodetector and Pearson correlation to identify and rank influencing factors.

First, the research team assembled an extensive dataset covering 1,037 large and hyperscale data centers across 29 Chinese provincial-level regions from 2016 to 2022. Spatial metrics—including the nearest-neighbor index, kernel density estimation, geographic concentration index, imbalance index, and Anselin Local Moran’s I—were computed using ArcGIS to characterize overall distribution, local clustering, and temporal change. They found the nearest-neighbor index consistently fell below 1 (R≈0.25 in 2022), which confirmed a statistically significant clustering pattern. Kernel-density mapping further exposed a persistent “dense east, sparse west” configuration anchored in three metropolitan cores: Beijing–Tianjin–Hebei, Yangtze River Delta, and Pearl River Delta. Over time, these point clusters expanded into belt-like corridors, while secondary concentrations emerged in Guizhou, Sichuan, and Inner Mongolia—evidence of a slow policy-driven westward diffusion. The authors found the imbalance index declined from 0.62 to 0.51 over the study period, which suggest gradual convergence among provinces though imbalance remained marked. Lorenz-curve analysis showed that by 2022, half of all data centers were still located in just five coastal provinces, however, the curve’s movement toward equality reflected early success of the EDWC strategy. Local Moran’s I results revealed “high-high” clusters in Shanghai and “high-low” contrasts in Guangdong, indicating that strong regional economies still exert gravitational pull, while adjacent provinces lag behind. Afterward, a combined approach of Geodetector and Pearson correlation analysis was used to identify and rank the drivers of data center distribution across the fourteen variables, which were categorized into seven key dimensions: policy environment, economic foundation, green transformation, market demand, cost control, operation and maintenance safety, and data center attributes. Computing demand (digital economy index) and economic development (per-capita GDP) exhibited the strongest national-level correlations with data-center density. However, from 2016 to 2022, the explanatory power of green-energy availability and energy-performance (PUE) rose markedly, while purely economic factors declined, revealing a paradigm shift toward environmental optimization. Regional contrasts were pronounced: in the east, market demand and GDP remained decisive; in the central regions, fiscal support dominated; and in the west, renewable-energy supply and policy incentives played leading roles. Interaction-detector tests demonstrated nonlinear enhancement between paired factors—particularly between fiscal support and carbon-emission level, and between fiscal support and PUE—underscoring the interdependence of economic policy and sustainability. Hypothesis testing further showed that renewable-energy access positively affected site density in technologically advanced provinces but still not in the under-electrified west, where infrastructure gaps impede utilization.

In conclusion, the research work of Professor Donglin Chen and colleagues contribute a rare, quantitative portrait of China’s data-center landscape, illustrating how spatial patterns mirror both economic gravity and ecological adaptation. The innovation of the work lies in integrating spatial heterogeneity analysis with multi-factor interaction detection, allowing non-linear relationships between economic, policy, and environmental variables to emerge with unprecedented clarity. The authors successfully bridge a methodological divide between descriptive mapping and influencing factor identification by fusing geostatistical metrics with Geodetector modeling, and enriched the analytical toolkit for spatial-energy research. Their findings confirm that the nation’s digital infrastructure remains concentrated in the east but is progressively realigning under government-led coordination. This transition embodies a broader policy evolution—from market-driven agglomeration to strategically guided, low-carbon regionalization. Additionally, the work highlights that spatial redistribution itself can function as an instrument of energy efficiency. New facilities concentrated in resource-abundant western regions reduces dependence on coal-fired grids and mitigates thermal-management loads through cooler climates and cheaper renewable electricity. The demonstrated strengthening of interactions among fiscal policy, renewable availability, and energy performance indicates that holistic governance—rather than isolated technological upgrades—yields the most substantial sustainability gains. The study also validates the Geodetector model’s suitability for analyzing high-dimensional, non-linear relationships typical of infrastructure geography, where multicollinearity often frustrates conventional econometrics. Moreover, the results provide evidence-based guidance for both policymakers and industry planners. For instance, central authorities differentiated regional strategies are essential: enhancing digital economy infrastructure in the west must proceed in tandem with grid modernization and data-transmission capacity, while in the east, stricter efficiency benchmarks and fiscal disincentives can curb excessive clustering. On the other hand, provincial governments can use localized renewable resources: solar in Ningxia, hydropower in Sichuan to attract greener data-center investment without replicating the east’s carbon footprint. Also for operators, the study highlights that future competitiveness will hinge on connectivity or cost as well as on alignment with sustainability metrics such as PUE and carbon intensity. Furthermore, the new study reframes data-center expansion as both an economic geography problem and an environmental governance challenge and shows that spatial balance and carbon neutrality can both happen: rational distribution guided by scientific metrics can reconcile growth with climate responsibility. Moreover, the integrated methodological framework—combining spatial-statistical diagnostics with Geodetector interaction analysis—can be replicated to study other energy-intensive digital infrastructures, from blockchain mining to cloud-computing clusters.