Chinese technology giants are rapidly increasing spending on artificial intelligence infrastructure but face greater financial constraints than their U.S. counterparts, whose larger and more profitable core businesses provide stronger cash flow and greater capacity to sustain heavy investment, Moody’s Ratings said.
Capital expenditure by leading Chinese technology companies is expected to rise to about $140 billion in 2026 and $165 billion in 2027, from $65 billion in 2025, the rating agency said in a report.
However, the six major U.S. hyperscalers are projected to spend more than $785 billion in 2026, roughly six times the combined total of their Chinese counterparts, with spending expected to approach $1 trillion in 2027.
The gap highlights the different financial positions of the two groups as competition intensifies to build the computing infrastructure needed to support increasingly sophisticated AI models and applications.
Alibaba Group Holding plans to invest RMB380 billion ($57 billion) over three years, a programme announced in February 2025. ByteDance is planning capital expenditure of up to $70 billion in 2026, according to a Bloomberg report cited by Moody’s.
Quarterly results also point to an acceleration in AI-related spending by Chinese companies. For the quarter ended June 2026, Alibaba reported RMB67.7 billion in capital expenditure, while Tencent Holdings’ spending rose 176% year-on-year to RMB52.8 billion. Baidu’s capital spending nearly tripled to RMB11.4 billion.
Despite the faster growth in Chinese spending, Moody’s said the difference in investment levels is expected to result in a widening gap in installed computing capacity.
U.S. data centre capacity reached 52 gigawatts at the end of 2025, compared with 28 GW in China, according to the International Energy Agency. The IEA expects U.S. capacity to reach 100 GW by 2030, while China’s capacity is forecast to expand to 67 GW.
Differences in infrastructure costs, labour costs and government support partly offset the disparity in capital spending.
Spending on data centre facilities, cooling systems and related infrastructure is significantly lower in China than in the United States on a per-megawatt basis, while lower land, construction and operating costs also improve deployment economics.
Government support further reduces the effective capital burden on Chinese technology companies, Moody’s said, with regional governments providing measures including land and green-energy access, infrastructure subsidies, faster project approvals and tax incentives.
Access to advanced hardware, however, remains China’s biggest constraint. Restrictions on access to leading-edge chips from Nvidia have pushed Chinese hyperscalers and state-linked operators towards domestic alternatives.
Although China’s domestic chips and software ecosystem continues to improve, it still lags Nvidia’s technology and ecosystem, Moody’s said.
Chinese hyperscalers are also building AI businesses from a materially smaller revenue base than their U.S. peers. By contrast, U.S. hyperscalers benefit from highly profitable and cash-generative businesses outside cloud computing, providing greater capacity to absorb elevated capital expenditure.
Moody’s expects AI investment to weaken free cash flow and push leverage higher on both sides of the Pacific, with several companies expected to turn free-cash-flow negative in 2026 and 2027 as capital expenditure outpaces growth in operating cash flow.
Still, substantial cash reserves remain an important credit strength for both groups, providing capacity to absorb 12 to 24 months of elevated capital expenditure without a material deterioration in credit quality, Moody’s said.