Despite their strikingly different approaches to developing and deploying artificial intelligence, one aspect is common across China and the United States: data and computing centres are proliferating in both countries, especially in remote locations.
While the US is seeing residents and localities fight back against plans for data centres in their backyards, regional officials in China are embracing them.
In fact, in China there is a nationwide scramble for such facilities – considered the backbone of AI – with some major cities losing out to towns and counties when it comes to new investments in the field.
While the speed and scale of the buildout align with Beijing’s overarching tech push, some experts are concerned that an overzealous data centre race may lead to overbuilt infrastructure.
In Shatian, a small coastal town in Dongguan city in southern Guangdong province, municipal authorities announced an aggressive plan earlier this month to invest 100 billion yuan (US$14.7 billion) to build a 9,000-acre (3,600-hectare) data and computing park on a barren island, with scheduled completion by 2030.
In eastern Zhejiang province, the little-known county of Jiashan bordering Shanghai has opened the first phase of a 5 billion yuan data centre with a capacity of 15,000 petaflops per second, a unit equivalent to 1 quadrillion calculations. When fully completed, the total capacity will double.
The data centre, designed for training next-generation AI models as well as molecular and drug development, would process data faster than traditional supercomputers in the US, according to Zhejiang media reports, citing local officials.
Jiashan county already hosts Alibaba Group Holding’s Yangtze River Delta Intelligent Computing Base and a computing centre operated by China Telecom. Alibaba owns the South China Morning Post.
“Previously, computing power clusters of this scale would only be found in top-tier cities like Shanghai and Shenzhen ... Now, Jiashan has a growing cluster of them,” Zhejiang Daily quoted local officials as saying.
Yangzhou, a city in Jiangsu province, has also broken ground on a hyperscale data park costing 60 billion yuan to build and boasting a projected capacity of 30,000 petaflops, with the backing of investors including Tencent Holdings and China Telecom.
Similar facilities of varying scale are also sprouting up in central and western regions like Hunan and the Xinjiang Uygur autonomous region. In Ulanqab, Inner Mongolia, a 30 billion yuan computing centre is taking shape.
However, Tang Dajie, a senior researcher at the China Enterprise Institute think tank, warned of “irrational” investments in the sector.
“AI is on everyone’s lips, so a headlong rush into related projects is almost inevitable. Local governments, driven by tenure performance targets, often exhibit irrationality,” Tang said.
He pointed to the stark contrast seen in the US, when some state governments and residents are pushing back on data centres in their towns.
States like New York, Virginia and Ohio have either moved to bar state agencies from issuing new permits or rolled back support policies over grid reliability and carbon emission targets.
“AI and data centre projects in China are in part cooked up by local officials, while the buildout in the US is more driven by tech firms, with local authorities usually scrutinising projects,” Tang said.
China’s need for more AI infrastructure was justified and Beijing could step up its oversight and coordination, he added.
“It still makes sense to build facilities outside cities, where electricity costs are low and there is abundant land or a favourable climate ... China already has a national policy for data processing in western provinces to serve the need in coastal regions,” Tang said.
In 2022, China embarked on the “east data, west computing” project, whereby data centres in China’s western inland would process the computing workload generated in its more densely populated eastern coast areas.
Xu Jingbin, a researcher with the Zhejiang Industry and Information Technology Research Institute, said that despite the risks of overbuilding facilities, the faster and cheaper development of AI infrastructure could help China gain an advantage against the US.
“As generative AI surges, demand for computing power is growing exponentially. Whoever controls low-cost, high-performance computing power holds a big edge ... This applies to the tech competition between countries,” Xu was quoted as saying by Zhejiang media.
“China usually has the synergy, from decision-making to implementation and between localities and tech firms. This is something still lacking in the US.” -- SOUTH CHINA MORNING POST
