China’s ambitious goal to green its rapidly expanding artificial intelligence infrastructure is colliding with the practical realities of power grid management, creating a paradox that threatens to complicate the country’s climate targets. The core issue lies in the fundamental mismatch between the intermittent nature of renewable energy sources such as wind and solar and the relentless, 24/7 power demands of AI data centers, which cannot tolerate even brief interruptions in supply.
Beijing has set an explicit, aggressive target: authorities aim to have renewable energy supply approximately 80% of the AI data center sector’s total power consumption by 2030. This represents a massive leap from the estimated 11% share in 2023. The 2026 government work report further emphasized this priority, mandating tighter integration between computing infrastructure and clean power supply, and a green-data-center action plan requires new projects in the national computing hubs to source most of their electricity from clean sources.
The Scale of the Incoming Demand
The urgency of the problem stems from the sheer scale of China’s AI buildout. Power demand from Chinese data centers is projected to surge by 300 billion to 500 billion kilowatt-hours between 2026 and 2030, an increase that would account for nearly a fifth of the country’s total electricity demand growth over that period.
Installed data center capacity is on track to reach roughly 40 gigawatts by the end of 2026, up from 32 gigawatts at the end of 2025, with some forecasts projecting the sector will surpass 60 gigawatts by 2030. The sheer volume of additional power required is enormous, and the buildout is accelerating precisely when grid operators are under the most pressure to decarbonize.
Intermittency vs. Reliability
The primary hurdle is not a lack of renewable generation capacity. China added hundreds of gigawatts of solar and wind in 2025 alone, and wind, solar, and storage are expected to overtake coal as the largest source of generation before the end of the decade. The problem is not a shortage of clean electrons overall, but the inability to guarantee them at the specific time and location where an AI cluster needs them — without any tolerance for interruptions.
AI training runs require a massive, steady power draw. The load profile of an AI training run looks nothing like the steady hum of a traditional data center, and grid operators are wary of taking on the added risk of balancing intermittent supply against a customer that cannot tolerate interruption. Faced with a choice between intermittent clean power and reliable conventional power, grid operators have consistently chosen reliability, and in China, reliable power still means coal.
Green Certificates as a Workaround
To meet green targets on paper, many operators have turned to green electricity certificates, which allow them to purchase the environmental attributes of clean power without a direct physical connection to a renewable source. This approach meets regulatory metrics and allows companies to claim green credentials, but it does not necessarily change which electrons power the servers.
The tension is not unique to China. Grid operators worldwide are being asked to absorb AI loads faster than they can build them, from Denmark pausing new data center connections to the EU asking households to ease off during peak hours. What distinguishes China’s situation is the combination of scale, speed, and the ambition of its targets, which is precisely why the gap between the stated goal and the current grid reality is worth watching closely.
Solving this paradox will require significant advancements in energy storage technologies, grid management systems, and, potentially, new models for colocating data centers with dedicated renewable generation assets. Until those solutions mature, China’s AI infrastructure will continue to run, at least in part, on coal.
Policy Ambition vs. Grid Reality
The gap between China’s stated renewable targets and the current reality of its data center power mix is not simply a technical problem—it is also a governance challenge. The entities responsible for building data centers (technology companies and state-owned enterprises) operate under different incentive structures than the entities responsible for managing the power grid (regional grid operators and the National Energy Administration).
Aligning these incentives so that clean power procurement becomes the default rather than the exception will require regulatory mechanisms that go beyond simple mandates. Some analysts have proposed that China consider requiring new data center projects to physically co-locate with dedicated renewable generation capacity, rather than relying on certificates. Others have suggested that AI companies should be required to publish their actual energy mix, not just their certificate-adjusted figures.
How Beijing resolves this governance challenge will have significant implications not only for its own climate targets but also for the global AI industry’s carbon footprint, given China’s central role in AI hardware manufacturing and model training.