China's factory PMI up to 50.1 – first growth in three months
China's industry returned to growth in September, and the figures arrive just as analysts try to measure how large the AI investment boom really is. But the link between the two is, for now, attributed rather than documented.
China's official purchasing managers' index for manufacturing (PMI) rose to 50.1 in September, up from 49.8 in August, according to the National Bureau of Statistics. It is the first time in three months that the measure has been above the 50-point threshold separating growth from contraction. A private survey, the RatingDog PMI, supports the picture: it rose to 52.1 from 51.5 (Associated Press).
What the figures actually show
The subindices reveal a nuanced picture. Production rose to 51.7 from 50.4 — a clear improvement. But new orders slipped slightly to 50.5 from 50.6, and new export orders fell to 50.0 from 50.1, in other words practically flat (AP).
The export-order detail matters for interpreting the story. Export orders measure future bookings, not shipments already on their way out of the factory gates. A narrative that the AI boom is lifting Chinese exports should ideally also be reflected in future orders — and there, the curve is flat.
The AI link is attributed, not documented
AP's reporting nevertheless ties the factory rebound to the global AI boom: "While China's economy is under increasing pressure from a downturn, with sluggish domestic demand and investment, partly hit by weaknesses in the property sector, the global AI boom has lifted Chinese exports of high-tech goods," AP writes (AP).
It is worth emphasizing what this is: an interpretation from a news source, not something that can be documented directly from the official figures. The statistics bureau's PMI report contains no data broken down by export categories, and none of the available sources can quantify how much of the September upturn is actually due to AI-related demand. One document — an AP quote that breaks off mid-argument — suggests that more analysis lies behind it, but what is missing cannot be reconstructed.
How big is the AI spending boom? Goldman Sachs' estimate
Independent of China's PMI, analysts are trying to put a figure on the investment boom itself. Goldman Sachs Global Institute and Goldman Sachs Global Investment Research estimate that the five largest hyperscalers will spend around $800 billion on capital expenditures in 2026, an increase of 94 percent from 2025. According to the estimate, spending will continue to rise — to $1.4 trillion by 2028, and to around $1.6 trillion annually by 2031. In total, Goldman Sachs estimates that roughly $7.6 trillion will be invested in AI infrastructure — computing, data centers and power supply — in the period 2026 to 2031 (24/7 Wall St. via Yahoo Finance).
The Goldman analysis also points to what the demand is doing to suppliers: constrained access has, according to the analysis, pushed memory chip producers' gross margins toward 80 percent, more than double the historical average — a sign of significant pricing power.
It is worth keeping the figures on the right side of the line between observation and projection: these are Goldman Sachs' own forecasts, conveyed through secondary sources, not recorded spending. Forward-looking estimates about a phenomenon as new as AI infrastructure involve considerable uncertainty — both about the volume itself and about how the investments will actually play out in employment, production and commodity demand.
Markets reacted immediately
Regardless of how one interprets the AI link, markets traded on the figures. The news that China's factory activity grew for the first time since June lifted mining stocks listed in London — China is a major consumer of industrial metals — and contributed to gains in Hong Kong, Shanghai and Tokyo, according to AFP (AFP via MSN).
The picture of a global investment boom received further support the same week: the US Department of Commerce revised its estimate of second-quarter GDP growth up by 0.7 percentage points to 2.2 percent, and analysts — including Arun Sundaram of CFRA Research — pointed to the AI investment boom as one of the drivers. According to the AFP report, part of the elevated bond yields is also attributed to the AI boom (AFP via MSN).
What remains as open questions
Several unanswered questions remain before one can say that China's factory rebound is due to AI.
First: How much of the September upturn is actually due to AI-related exports, as opposed to domestic stimulus measures, seasonal variation or a low comparison base after three weak months? None of the available sources separates these factors from one another.
Second: the export orders indicator, which stood at exactly 50.0 — neither growth nor contraction — provides no support in itself for a narrative of an export-driven upturn. Stronger evidence would require data on export categories or business surveys distinguishing high tech from traditional industry, but such data is not included in the official figures.
Third: Goldman Sachs' $7.6 trillion is an estimate based on models of how computing demand will develop. If the projections over- or undershoot, the entire story of an "AI-driven" economic effect — in both China and the US — will change accordingly.
What can be established, however, is this: for the first time in three months, China's factory data show growth, a private measure confirms the direction, and markets reacted immediately. And at roughly the same time, both AP reporting and Goldman Sachs analyses point in the same direction: that the global AI spending boom has grown large enough to leave traces in macroeconomic data. How deep those traces are remains to be seen.