"When it comes to projects, there are a few things – artificial intelligence, computing power, and new energy vehicles. Do all provinces in the country have to develop industries in these directions?"
Eastern Data, Western Computing: lofty targets, measured results
The Three-year Data Center Plan and the Eastern Data, Western Computing (EDWC) initiative set explicit, ambitious goals: a 20 percent annual growth rate for server rack installation, a 60 percent increase in utilisation rates, and more than 200 eflops of national computing power. By 2024, reporting shows 633 hyperscale and large data centres were constructed under EDWC, bringing China’s computing power to 268 eflops—above the plan’s 200 eflop benchmark on paper. Yet those headline figures mask a widening gap between installed capacity and productive use.
Cool climate and cheap power: the appeal of the west
China’s strategy moved energy-intensive tasks westward to exploit colder climates and abundant renewable energy. Distributed solar projects in the west can bid as low as 0.19 yuan per kilowatt-hour versus up to 0.43 yuan per kilowatt-hour in some eastern regions. Provinces such as Guizhou, Gansu, and Ningxia became attractive locations for data-centre construction. Even provinces not originally designated as EDWC clusters — notably Shaanxi — rapidly expanded capacity; Shaanxi constructed 22 large-scale data centres and three big data industrial parks, backed by millions of yuan in subsidies.
Provincial competition, subsidies and overinvestment
Local governments fueled a construction frenzy with subsidies, rent reductions, tax benefits and other incentives to boost regional GDP. Provinces including Shandong and Guizhou enacted provincial-level expansion plans; compute-voucher programs were rolled out nationwide in 2023 to subsidise AI development, with Beijing, Shanghai and Chengdu offering more favourable incentives than western provinces such as Kunming. The result was redundant investment and oversupplied infrastructure.
Guizhou, once presented as a poster child for EDWC, illustrates the downside: despite heavy incentives and subsidies the province has not turned a profit on its data-centre expansion, now ranks 22nd in regional GDP, is burdened by high debts, and faces corruption issues in its big-data industry.
Transmission, latency and utilisation: where the plan frays
EDWC’s model assumed western centres would take bulk, energy-heavy workloads while eastern centres handled low-latency inference for real-time users. In practice, structural barriers emerged. Remote western sites often lack the fibre-optic capacity needed to move massive data volumes in real time, forcing operators to spend more on transmission. Western regions struggle to attract skilled workers and local customers. Because AI inference requires low-latency networks and eastern regions contain more “hot data” and real-time users, eastern data tends to remain processed in the east.
The mismatch shows in utilisation figures: some western data centres report utilisation as low as 20–30 percent, far below the policy goal of more than 60 percent. Renewable-energy output reduction rates in the western region continue to exceed 30 percent, and eastern data centres in Shanghai, Beijing and Shenzhen strain grids that rely heavily on coal. Beijing has since moved to consolidate capacity, restating that all data centres should have utilisation rates of no less than 60 percent and that no new large or super-large data centres should be built in cities where existing centres operate below 50 percent.
Chips, subsidies and the risk of stranded assets
Reports of overcapacity prompted discussions of selling excess computing power and revealed fragmentation in chip system design and standards. In November 2025, Beijing introduced electricity subsidies of up to 50 percent for data centres that used domestically produced semiconductors—policy action that followed restrictions preventing Chinese technology giants from purchasing Nvidia’s most advanced chips. The subsidies were framed to offset the added energy costs of using less advanced chips and to boost utilisation, and they are cited as evidence of weak demand relative to EDWC expectations.
Yet fragmentation in semiconductor architectures, a shortage of skilled operators, and mismatches between provincial incentives and workload distribution suggest many facilities could become stranded assets: impressive on paper but underutilised in practice.
What this means for policymakers, technologists, and provincial governments
- Policymakers and regulators: consolidation measures and utilisation thresholds signal a shift from expansion to optimisation; they will need to reconcile provincial incentives with national coordination to reduce redundancies and debt exposure.
- Technologists and operators: transmission bottlenecks, heterogeneous chip systems and low local demand mean engineering and operations teams must prioritise interoperability, efficient scheduling, and workforce development to raise utilisation.
- Provincial governments and investors: those that expanded rapidly face fiscal stress and potential asset stranding; provinces such as Guizhou highlight the political and economic costs when local ambitions outpace feasible demand and logistics.
EDWC transformed China’s geography of computing—creating vast, cheap-power zones in the west and dense, latency-sensitive nodes in the east—but the initiative’s planners did not fully close the loop between energy supply, network transmission and actual AI workloads. Beijing’s recent interventions—consolidation directives and targeted electricity subsidies—address symptoms but not all structural causes. The remaining question is whether tighter coordination, improved transmission, and standardised systems will be enough to convert installed eflops into sustained, productive computing capacity rather than a nation of impressive yet idle machines.
