Within the race towards China for AI supremacy, the USA dominates relating to entry to essentially the most cutting-edge semiconductors.
However relating to powering the large knowledge centres that run on AI chips, China holds the clear benefit.
Really helpful Tales
checklist of 4 objectsfinish of checklist
That’s as a result of knowledge centres, the sprawling computing services wanted to coach and run AI fashions, require huge quantities of power.
A typical knowledge centre can devour as a lot electrical energy as 100,000 households, whereas next-generation “hyperscale” services can gobble up as a lot energy as two million properties, in accordance with the Worldwide Power Company (IEA).
China’s entry to an plentiful provide of low-cost electrical energy locations it within the best place to satisfy such colossal power calls for.
China already generates greater than twice as a lot electrical energy because the US, a lead that’s anticipated to widen amid an aggressive state-led funding within the nation’s power grid.
BloombergNEF, a analysis supplier, estimates that China will add greater than six occasions as a lot electrical energy technology capability because the US over the subsequent 5 years.
A lot of that additional capability will be within the type of renewables corresponding to photo voltaic and wind.
In 2025 alone, China elevated its wind and solar energy capability by greater than 430 gigawatts, accounting for greater than half of the extra capability within the renewables added globally that yr.
A key ingredient of China’s AI technique entails integrating its knowledge centres into its quickly increasing renewables sector.
Beneath the “East Information, West Computing” initiative, China’s authorities is concentrating the development of recent knowledge centres within the nation’s sparsely populated inside, the place land and renewable power sources are plentiful in contrast with the closely built-up japanese seaboard.
Earlier this month, Beijing introduced the beginning of operations on the nation’s first “large-scale” renewable power undertaking to be linked immediately to an information centre.
The five hundred-megawatt wind and photo voltaic undertaking, positioned within the northwestern Ningxia area, will energy a cloud knowledge centre operated by China Datang via a “devoted transmission line”, China’s administration physique for state-owned enterprises stated in a press release on Could 12.
“In the long term, the nation that may present low-cost, steady, low-carbon electrical energy could have a serious benefit in AI infrastructure,” Qiyang Xiong, a PhD candidate at Renmin College of China who specialises in AI and power coverage, advised Al Jazeera.
“China is a world chief in photo voltaic, wind and ultra-high-voltage transmission,” Xiong stated.
“This offers it a bonus in supplying western knowledge centre clusters with massive volumes of comparatively low-cost, clear electrical energy.”
Narrowing the hole
For now, the US nonetheless has the biggest knowledge centre footprint by a large margin.
In accordance with Stanford College’s AI Index, the US had an estimated 5,427 knowledge centres in 2025, in contrast with 449 in China.
The US accounted for 45 p.c of the 415 terawatt-hours of electrical energy consumed by knowledge centres in 2024, adopted by China and Europe with 25 p.c and 15 p.c, respectively, in accordance with the IEA.
In 2026 alone, Silicon Valley’s Amazon, Microsoft, Meta and Alphabet are projected by Morgan Stanley to spend $630bn on knowledge centres and different AI-related funding, vastly greater than Chinese language tech giants corresponding to Alibaba, Tencent and ByteDance.
However as China constructs knowledge centres at a blistering tempo – its variety of knowledge centre racks grew 30 p.c yearly from 2016 to 2023, in accordance with the China Academy of Info and Communications Expertise – the hole between the superpowers is quickly narrowing.
Dealing with US export controls on top-end Nvidia chips manufactured by the Taiwan Semiconductor Manufacturing Firm (TSMC), China has more and more turned to partially state-owned Semiconductor Manufacturing Worldwide Company (SMIC) to provide chips designed by native tech corporations corresponding to Huawei.
By 2030, China’s knowledge centre capability is predicted to succeed in 60 gigawatts, practically double its present stage, in accordance with an evaluation by Rystad Power, taking over 2.3 p.c of the nation’s complete electrical energy demand.
“China’s massive manufacturing base and fewer stringent regulatory atmosphere imply that the development of knowledge centres and supporting power infrastructure can occur much more quickly than within the US,” Leah Fahy, senior economist for China at Capital Economics, advised Al Jazeera.
“Modular Huawei knowledge centres can now be constructed in six months, whereas equivalents within the US take not less than a yr,” Fahy stated.
Energy grids below pressure
In the meantime, there are already indicators that the AI rollout within the US is bumping up towards energy constraints.
Power consultancy Wooden Mackenzie stated earlier this yr that the constraints of the US power grid had resulted in a 50 p.c quarter on quarter drop in new knowledge centre initiatives on the finish of 2025.
Technical limitations have been compounded by a rising backlash towards knowledge centres inside communities throughout the US – pushed partly by the pressure the services place on native grids – a problem not confronted by China, the place opposition to the federal government is closely restricted.
A minimum of 36 knowledge centres have been blocked or stalled within the US between Could 2024 and June 2025, in accordance with Information Middle Watch, a analysis undertaking by AI safety firm 10a Labs.
US tech leaders, together with Tesla’s Elon Musk, Nvidia’s Jensen Huang and OpenAI’s Sam Altman, have brazenly acknowledged China’s edge within the power area.
“The limiting issue for AI deployment is essentially electrical energy,” Musk stated in an interview on the World Financial Discussion board in January.
“Very quickly, possibly even later this yr, we’ll be producing extra chips than we are able to activate – apart from China. China’s progress in electrical energy is large.”
Advancing AI is now an “electrical energy downside as a lot as a chip downside”, stated Howard Yu, director of the Middle for Future Readiness at IMD Enterprise College in Lausanne, Switzerland.
“The winners of this cycle will personal the silicon, the facility contracts, and the cooling water, in that order, and China has constructed its technique across the enter it controls,” Yu advised Al Jazeera.
China’s power benefit will not be with out its personal limitations.
Regardless of Beijing’s push to meld its AI ambitions with the wind and photo voltaic sources of its distant western areas, most knowledge centres are nonetheless positioned in and round japanese megacities corresponding to Beijing, Tianjin, Shanghai, Guangzhou and Shenzhen.
“These locations additionally face energy provide difficulties and have launched restrictions on new knowledge centres,” Anders Hove, a senior analysis fellow on the Oxford Institute for Power Research, advised Al Jazeera.
Hove added that China’s energy grid additionally suffers from a excessive diploma of fragmentation that forestalls the seamless circulate of electrical energy between areas.
“China’s energy system is organised and dispatched primarily on the provincial stage, with transmission corridors appearing primarily as one-way energy flows,” Hove stated.
“Although the central authorities has known as for regional wholesale markets and extra granular buying and selling intervals, that is continuing slowly,” he added.
High quality management
Although fast, China’s knowledge centre rollout has additionally confronted high quality points, stated Kyle Chan, a analysis fellow on the Brookings Establishment who specialises in Chinese language tech and industrial coverage.
“They’re making an attempt to construct heterogeneous chip clusters that group collectively totally different {hardware} programs. This makes it more difficult to run AI workloads,” Chan advised Al Jazeera.
“There have been points with the construct high quality of some Chinese language knowledge centres, significantly when the developer doesn’t have correct expertise with such a fancy undertaking,” Chan stated.
China has additionally has some method to go to slim the hole between knowledge centre capability and utilisation, stated IMD Enterprise College’s Yu.
“Beijing’s personal estimates put it at 20 to 30 p.c, and even SMIC’s chief has warned the brand new capability may sit idle,” Yu stated.
“One method to body the entire race: the US has the chips and is brief on energy, whereas China has the facility and is brief on chips. Every is sprinting to repair its personal bottleneck.”