Opinion: U.S. AI investment growth may peak in Q4, with a decline in 2027

October 27: Shenwan Hongyuan released a macro research report stating that U.S. AI investment growth may peak around the fourth quarter of this year and enter a downturn phase in 2027; approximately 74% of global AI industry chain capital expenditure flows to data centers.

The report states that as of the second quarter of 2026, hyperscale cloud providers' capital expenditure accounted for 66% of the AI stack, making them the primary spenders. According to Bloomberg consensus estimates, the capital expenditure scale of the five major U.S. cloud providers (Amazon, Google, Meta, Oracle, Microsoft) in 2026 may exceed $800 billion, with a year-over-year growth rate of approximately 94%, accounting for about 2.5% of U.S. GDP; in 2027 and 2028, the scale may continue to expand, but growth rates may fall back to 40% and 10%, respectively. In terms of object structure, approximately 74% of global AI industry chain capital expenditure flows to data centers, of which about 55% is directed to servers/GPUs.

Shenwan Hongyuan, based on project-level data calculations, states that the average data center construction cycle is about 20 months, with an average shelving rate of about 33%; investment within one year is mainly determined by existing planned and under-construction projects. Under the three shelving rate scenarios of 20%, 33%, and 45%, the calculations all point to the peak of U.S. AI investment growth occurring in the fourth quarter of 2026, approximately one quarter later than the market consensus on the capital expenditure growth rate of the five major cloud providers, and entering a growth downturn phase in 2027 is likely a high-probability event.