Although US tech giants outspend their Chinese counterparts by a wide margin on AI infrastructure, new analysis shows China’s lower operational costs and preferential policies enable its firms to achieve much closer computing power capacity than spending alone suggests.
- US AI hardware spending outpaces China fivefold by 2027
- China’s cheaper power, land and subsidies boost compute efficiency
- Limited access to cutting-edge chips remains a key challenge
What happened
Moody’s Ratings analysis reveals that US technology giants, including Microsoft, Amazon Web Services, Alphabet, Meta Platforms, and Oracle, alongside AI cloud provider CoreWeave, are forecasted to invest nearly $1 trillion in AI compute infrastructure by 2027. In comparison, China’s major tech companies expect to double their capital expenditures to approximately $165 billion over the same period.
Why it matters
China’s ability to stretch each dollar of investment into greater compute capacity presents a formidable challenge to US tech dominance in AI infrastructure. Initiatives like Beijing’s “East Data, West Computing” strategically move power-intensive AI workloads to inland regions rich in renewable energy, enabling large energy savings on power and cooling costs that are critical in AI training workloads.
However, China’s compute cost edge largely does not extend to advanced semiconductor chips, where US firms lead due to restricted access to cutting-edge Nvidia processors. Chinese domestic chips tend to be less energy efficient, which offsets some of the cost advantages, but overall capacity gains signal that pure expenditure numbers only partly capture competitive positioning in AI compute.
What to watch next
Looking ahead, total data center capacity in the US is slated to reach 100 gigawatts by 2030 versus 67 gigawatts in China, narrowing the relative gap even as the US retains a lead. China’s faster growth rate in capacity, currently about 19% annually compared to 14% in the US, underscores a rapid catch-up in physical AI compute infrastructure despite chip constraints.
Future developments to monitor include China’s progress in boosting domestic semiconductor performance and efficiency, as well as potential shifts in geopolitical or trade policies affecting advanced chip access. The interplay between raw spending, capacity growth, and hardware capability will shape the evolving global AI compute race in the coming years.