China scales AI infrastructure using lower computing costs

Major Chinese technology companies are expanding their artificial intelligence (AI) computing infrastructure with significantly lower capital costs compared to their American counterparts. While U.S. hyperscale operators are projected to spend nearly $1 trillion by 2027, Moody's Ratings indicates that China's ability to secure cheaper land, power, and cooling infrastructure allows them to achieve competitive computing power with less capital.

China's strategy involves shifting AI training workloads to western and inland regions, such as Inner Mongolia, to leverage lower electricity and cooling costs. Although the U.S. is expected to maintain a lead in absolute data center capacity, China's expansion rate is projected to be higher, with an average annual growth of 19% compared to 14% in the U.S.

Despite these cost advantages, Chinese firms face significant hurdles, including restricted access to advanced Nvidia chips and a gap in power efficiency. Additionally, a substantial revenue gap remains between U.S. cloud leaders like Microsoft and Amazon and Chinese providers such as Alibaba and Baidu.