China’s energy regulators and technology ministries have moved from discussing the AI power challenge to legislating a response. On May 8, 2026, the National Energy Administration, the National Development and Reform Commission, the Ministry of Industry and Information Technology, and the National Data Administration jointly issued an action plan to promote what the government describes as the “mutual empowerment” of AI and energy infrastructure. The document targets transformational outcomes by 2030 and includes concrete milestones for 2027.
The plan is a direct response to a structural tension that China’s AI buildout has made acute: AI data centers consume substantial amounts of electricity, and that electricity is often generated from fossil fuels at a time when China is simultaneously trying to accelerate its renewable energy transition. Generating one million AI tokens requires between 0.1 and 0.3 kilowatt-hours of electricity, according to government analysis. At the scale China is operating and planning to operate, the aggregate demand number becomes significant quickly. Projected electricity demand for China’s data centers alone reaches 277 terawatt-hours by 2030 under current trajectories.
The 2027 and 2030 Targets
The near-term milestones are specific. By 2027, Chinese authorities plan to deploy five sector-specific large language models for the energy industry and establish more than ten “AI plus energy” pilot projects. The pilot framework will test the integration of AI systems with grid management, renewable energy forecasting, and smart energy infrastructure across selected sites before broader rollout.
By 2030, the plan targets a significant increase in the share of clean energy powering AI computing infrastructure, a formulation that deliberately links the data center expansion with China’s decarbonization commitments. The document also envisions what it calls “world-leading energy-specific AI capabilities” by that date, including breakthroughs in grid stabilization, demand forecasting, and renewable integration. That ambition goes beyond energy efficiency; it frames AI as an active tool for managing the energy system, not simply a consumer that must be managed within it.
The 2030 framing matters because it aligns the AI-energy initiative with China’s existing energy transition commitments, embedding AI infrastructure planning within the climate policy architecture rather than treating it as a separate concern. Brookings has noted that the energy demands of the AI race are creating strategic infrastructure questions for both the US and China; Beijing’s response is to treat those questions as an integrated planning problem rather than a technology-sector afterthought.
The Electricity Cost Advantage
China’s electricity prices, which range from approximately 0.3 to 0.7 yuan per kilowatt-hour depending on region and time of use , give its AI industry a meaningful cost advantage relative to US and European operators paying higher industrial power rates. Data center electricity costs are a significant component of AI inference and training economics, and China’s price differential translates directly into lower per-token operating costs at scale.
That advantage is part of why China’s government is now treating AI compute infrastructure and energy infrastructure as a single policy domain. China’s data centers are beginning to participate directly in electricity spot trading markets, an innovation that allows large computing facilities to buy power when renewable generation is abundant and prices are low , improving economics and grid stability simultaneously. The May 8 action plan creates a regulatory framework that would systematize and scale that kind of coordinated optimization across a much larger share of the country’s AI computing capacity.
The plan also addresses the “Eastern Data, Western Computing” strategy that has been moving AI workloads toward western China’s less-expensive power and land, while ensuring that data generated in eastern economic centers can be processed at lower cost elsewhere. Effective AI-energy integration requires coordination across fragmented electricity, computing, and eventually carbon markets , the action plan acknowledges this directly by involving multiple regulatory bodies with different jurisdictional mandates rather than assigning the problem to a single ministry.
China has separately been building out its data center capacity with aggressive speed. The action plan is the government’s formal attempt to ensure that the grid can meet the resulting demand cleanly and reliably, rather than letting AI expansion and energy transition work at cross purposes. Whether the coordination among four government bodies produces coherent implementation will be a key question as 2027 milestones approach. Historical precedent in Chinese industrial policy suggests that cross-ministry plans of this type can lose momentum at the implementation stage; the specific 2027 milestones, with their quantified targets, create at least the conditions for measuring whether that happens here.