China's "25% Export Boom" — Read the Denominator Before You Believe the Headline

Before anyone on this desk repeats the phrase "China's trade data smashed expectations" in a client call today, read this memo. Then read it again, because half of what's being reported as volume is actually price, and nobody upstream seems to be flagging the difference.

The headline: Chinese exports rose 25.0% year-on-year in August, up from 23.9% in July, roughly matching the Reuters poll median. Imports climbed 28.2%. The trade surplus widened to $119.09 billion, up from $112.5 billion in July, putting the year-to-date cumulative surplus at $805.51 billion — on pace to rival or beat last year's record $1.2 trillion. Semiconductor exports alone surged 129.8% year-on-year. Auto exports were up 43%. This is the data point everyone is using to argue China has "weathered" the Iran war and tariff pressure better than expected, that AI-driven demand is bailing out a soft domestic economy, that the growth machine is intact.

Now the part nobody's including in the same sentence: semiconductor and electronics prices are up as much as 700% year-on-year in places due to the global chip shortage. S&P Global's own estimate for actual port volume growth this year is roughly 5%. Five percent. Not twenty-five. The gap between a 25% value print and a 5% volume estimate isn't rounding error — it's the entire story, and it's being told backwards. China isn't shipping dramatically more stuff. It's shipping roughly the same amount of stuff at radically inflated prices, because the AI capex supercycle has turned semiconductors into a scarcity asset, and every dollar of that price inflation flows straight into the trade-surplus line as if it were organic export strength.

This matters for three separate reasons, and each one gets uglier the longer you sit with it.

First, the composition. Exports to the U.S. jumped 34.4% to $42.5 billion, pushing the bilateral surplus to $29.18 billion — up 44% year-on-year and the widest gap since Trump returned to office in January 2025. Exports to the EU, by contrast, grew just 6.7%, the slowest pace in ten months. That divergence lands three weeks before an expected Xi-Trump summit in late September, on data that will be read in Washington as evidence China is exploiting the AI chip bottleneck to run up exactly the kind of lopsided surplus the last trade fight was supposed to fix. Nobody in that meeting is going to caveat the number with "adjusted for a 700% price effect in semiconductors." They're going to see $29 billion and a 44% YoY jump and treat it as leverage, or as provocation, depending which side of the table they're sitting on.

Second, the domestic-demand read. Imports climbed 28.2%, and the bulls are citing that as proof household demand isn't collapsing. It might not be. But import values are subject to the exact same commodity and chip-price inflation distorting the export side — China is a massive net importer of the same semiconductors it re-exports as finished electronics, so a chunk of that 28.2% is China paying more for the same input volumes, not buying more of them. Reading rising import values as rising domestic demand in an environment of a documented 700% price spike in the relevant category is not analysis. It's taking the headline at face value because the headline is convenient.

Third, and this is the one that should actually worry anyone long the AI infrastructure trade: if chip scarcity pricing is inflating trade data by this much, the same scarcity pricing is inflating capex line items everywhere else in the chain simultaneously. Hyperscaler capex guides, foundry revenue, memory pricing, and now sovereign trade surpluses are all being pumped by the identical shortage. That's not four independent data points confirming a supercycle. That's one input constraint showing up four times in four different reports, and everyone treating it as four separate pieces of evidence because it arrived in four separate press releases.

None of this means the underlying AI infrastructure demand is fake. Semiconductor scarcity is real, chip prices are up because actual buyers actually want the actual chips, and China's positioning in that value chain is a genuine structural shift, not a mirage. What's fake — or at minimum badly mislabeled — is the idea that a 25% export print represents 25% more Chinese economic output hitting the water. It represents roughly 5% more stuff moving through the ports at prices that have exploded because the world is short a component everyone needs at once.

Desk recommendation: strip price effects out of every China trade release before it goes in front of a client this quarter, and flag the U.S. bilateral number specifically ahead of the Xi-Trump meeting — that $29 billion figure is going to get cited in Washington as if it were a policy failure rather than a chip-shortage artifact, and whoever's on that call should walk in knowing the denominator before someone else's spin fills the gap.