China’s AI race: Can it make energy and computing work as one?

China’s artificial intelligence (AI) pursuit is rapidly becoming an energy race. A country that develops the best AI models will gain little if it cannot provide enough reliable and affordable electricity to run them. China has recognised this and is increasingly treating computing power, electricity generation and the national grid as parts of the same strategic infrastructure.

‘East Data, West Computing’ programme

China’s AI expansion is creating enormous electricity demand. Data centres already consume a growing share of national power, and demand will rise sharply as larger models, cloud services and AI applications spread throughout the economy. Beijing’s response is not simply to build more power plants. It is attempting to coordinate AI infrastructure with renewable generation, energy storage, transmission and increasingly a nationwide computing network.

The most ambitious example is China’s “East Data, West Computing” programme. Launched nationally in 2022, it was based on a compelling geographical imbalance. Most Chinese internet users, tech companies and computing demand are concentrated in the east, while much of the country’s wind, solar and hydro resources, cheaper land and cooler climate are found further west. China therefore encouraged the building of large data centres in places such as Inner Mongolia, Gansu, Ningxia and Guizhou.

China’s 15th Five-Year Plan for 2026-2030 explicitly called for further advancing the East Data, West Computing initiative while developing this larger network. By March 2026, more than 80% of China’s intelligent computing capacity was located within the eight national computing hubs established under the programme. Beijing is now trying to connect these facilities through high-speed networks and common scheduling platforms so that computing tasks can be allocated across the country — similar to how electricity is dispatched through a power grid. 

Their advantages are substantial. Renewable electricity can be abundant and relatively inexpensive, while cheaper land and cooler temperatures in northern and high-altitude areas can reduce cooling requirements. Rather than transmitting every additional kilowatt-hour eastward, China can effectively transport the economic value of that electricity by using it to power data centres in the west, processing data there and sending the results back east.

Towards geographically flexible computing

But geography also creates disadvantages. Data must travel long distances through fibre networks, making network quality and latency critical. Western China is not one homogeneous environment either. Some regions are cold but dry; others have plentiful hydropower but warmer or more humid conditions. A location excellent for electricity may be poor for water, telecommunications, skilled labour, or proximity to customers.

There is another potential weakness in the strategy: building capacity is easier than using it efficiently. China’s centralised system can mobilise capital and infrastructure extraordinarily quickly, but this strength can produce overbuilding. Earlier official figures showed that utilisation of the ten national data centre clusters that sit within the eight national computing hubs was only about 63%, although it had been improving. New policies increasingly emphasise utilisation rates as well as construction, suggesting that Beijing itself recognises this problem.

This is why building a “national computing power network” may ultimately be more important than the East Data, West Computing project itself. The real objective should not be to move servers in the west but to make computing geographically flexible. AI training could follow cheap renewable electricity; urgent inference could remain near users; non-urgent workloads could move across regions and even across time according to electricity availability. In this model, computing becomes another form of flexible demand on the electricity system.

A smarter energy infrastructure powered by AI?

China is also reversing the relationship between AI and energy. Instead of merely asking how electricity can support AI, Beijing wants AI to improve the electricity system. Its “AI+Energy” strategy seeks to use AI to forecast renewable production and electricity demand, identify equipment failures, optimise battery charging and assist grid planning, dispatch and maintenance.

Grid operators therefore need a strong human-in-the-loop principle. AI should analyse enormous quantities of information and recommend actions, but humans must retain authority over critical decisions. Cybersecurity, independent verification and the ability to operate essential systems when AI fails will become increasingly important.

China has all the tools. Can it work?

There is also a paradox. AI consumes electricity, but AI can make electricity systems more efficient. More efficient grids can then support more AI infrastructure, which consumes still more electricity. Efficiency gains may therefore be overwhelmed by rapid growth in computing demand, a classic rebound effect.

China and the West face the same problem but approach it differently. The International Energy Agency expects global data centre electricity consumption to rise dramatically towards 2030, with the US and China responsible for much of the increase. America remains extraordinarily strong in advanced AI models, chips and private investment, but its electricity system is more fragmented. Technology companies can build data centres rapidly, while new power plants and transmission lines can take years because of regulatory, grid-connection and community constraints.

China’s strength is coordination. The government can simultaneously promote renewable generation, ultra-high-voltage transmission, energy storage, fibre networks and data centre construction. Its manufacturing capacity in solar panels, batteries and electrical equipment reinforces this advantage.

Yet the success of the East Data, West Computing initiative should not be judged by how many data centres appear in western provinces. Its real test is whether China can keep them highly utilised, connect them efficiently with eastern demand, match computing workloads with renewable electricity and avoid creating new water and environmental stresses.

Seen this way, the programme is broadly on the right track, but its evolution is more important than its original design. “East Data, West Computing” was essentially a geographical strategy. The emerging national integrated computing power network is becoming a resource-allocation strategy. That is a much more sophisticated idea.

Infrastructure constraining AI

China’s experiment is important because it is attempting to merge three national networks: computing, telecommunications and electricity. Renewable availability could increasingly determine where and when computing occurs, while AI could help manage the energy system supporting that computing.

If China succeeds, this infrastructure could become a significant competitive advantage. If poorly managed, the same system could create stranded computing capacity, grid stress, water shortages, inefficient investment and new cybersecurity vulnerabilities.

The global AI race is therefore no longer simply about algorithms and semiconductors. China and the West are discovering that AI is ultimately constrained by physical infrastructure. The winners will not necessarily be those that build the most data centres, but those that convert energy, water and computing infrastructure into useful intelligence more efficiently, sustainably and securely.

Related: China tech firms drive AI data centre boom in Southeast Asia | Can China win the AI race with cheap power?