China Is Making Digital Growth Answer for Its Own Energy Use

China’s new plan for coordinated digital and green transformation begins with a useful refusal: digital technology cannot be treated only as a tool for cutting emissions elsewhere. Computing facilities, communications networks and artificial-intelligence models have their own energy and material demands. A credible digital-green strategy must manage both sides.

The 2026–2030 plan was issued by seven bodies spanning cyberspace, economic planning, industry, environment, housing and urban development, agriculture and commerce. It contains 14 tasks across technology, infrastructure, industry, trade, consumption, city operations, environmental oversight and carbon management.

Its most consequential move is to place computing and electricity in the same planning frame. By 2030, the plan seeks major efficiency gains for computing facilities and 5G base stations. It calls for renewable-electricity consumption in key computing facilities to reach the renewable-consumption responsibility level of their province. It also promotes direct green-power connections, green certificates, long-duration storage and the movement of computing loads toward areas rich in wind, solar and hydropower.

Efficiency reaches inside the model

The plan does not confine efficiency to cooling or buildings. It reaches into the technical stack: integrated computing and memory, data compression, distributed training, high-bandwidth memory, low-power chips, mixture-of-experts models, dynamic sparsity, quantisation and pruning. It also proposes a lifecycle resource-consumption assessment for generative AI and encourages service providers to disclose how large models use resources.

That breadth matters. A more efficient data center can still consume more electricity if model scale and demand grow faster than efficiency improves. A leaner model can still depend on hardware whose manufacture, replacement and cooling carry substantial resource costs. Measuring only facility-level energy efficiency would miss those rebound effects.

The plan’s disclosure language is therefore an important opening, but it leaves design work ahead. “Resource use” can mean electricity, peak demand, water, hardware utilisation, embodied carbon or several of them. Comparisons will be weak unless reporting establishes functional units: per training run, per inference, per user task, per unit of useful compute or across the full service lifecycle. The eventual assessment method will determine whether disclosure changes decisions or merely adds another aggregate number.

Data must cross administrative boundaries

The second half of the plan treats digital infrastructure as an instrument of environmental management. It calls for linked data on greenhouse-gas assessment, environmental monitoring, energy use, carbon emissions and product footprints. Industrial parks are encouraged to build energy-and-carbon management centers. Environmental authorities are directed toward cross-checking zoning, impact assessment, permits and enforcement data with AI models.

This is not primarily a sensor problem. The difficult work is whether agencies and firms use compatible definitions, identifiers, time periods and quality controls. A model can find apparent inconsistencies that come from different reporting boundaries rather than misconduct. Conversely, clean-looking dashboards can conceal bad source data if organizations optimize what they submit rather than what they operate.

The plan recognizes that challenge indirectly by calling for stronger carbon-market infrastructure, better emissions factors, product-footprint databases, data-security protection and regular monitoring reports and development indices. Those elements create a possible chain from measurement to supervision. They do not guarantee that the chain will withstand commercial incentives, local variation or uneven administrative capacity.

The city becomes a control surface

At urban scale, the plan promotes city-information models, integrated operations platforms and digital twins for electricity, gas, heat and water. Distributed solar and storage would be brought into coordinated dispatch, while transport data would support safer and more efficient network operation.

The promise is attractive: fragmented assets become visible enough to operate together. The risk is that visibility is mistaken for control. A city platform cannot dispatch a privately owned battery without a rule or contract. It cannot correct a leaking water network without maintenance crews and capital. It cannot make agencies cooperate merely by placing their data on one screen.

China’s plan is strongest when it specifies the connection between digital tools and physical decisions: power-location choices for computing, energy-aware network operation, carbon measurement tied to management, and environmental data tied to permits and enforcement. Its weakest points are where outcomes are expressed as broad improvements without a common baseline.

The promised monitoring system, periodic reports and development index will therefore be central. They should reveal whether digital infrastructure is shifting toward cleaner power, whether absolute resource use is changing, and whether data integration is improving real industrial and urban performance. Without that double account, digitalisation can look green because it measures other sectors more closely while leaving its own footprint outside the frame.