China's AI boomtowns race ahead as rest of economy lags

Hefei's factories cannot keep up. That is the through-line of a new Bloomberg report on how China's AI build-out is producing a split-screen economy, with a handful of tech hubs delivering their biggest share of national growth in at least two decades in the first half, while the rest of the country slows to the bottom edge of Beijing's full-year target. The numbers are Nomura's.

The contrast Bloomberg draws is sharp. In Hefei, the memory-chip capital that lured CXMT with state investment funds, output can barely keep up with the world's appetite for AI hardware. In Changchun, the rust-belt home of gasoline-powered carmaking, officials admit to 'unprecedented' difficulties. One city is a magnet for AI capex, the other is watching an older industrial model exhaust itself in real time.

The awkward part for policymakers is how little of the AI windfall is reaching households. Bloomberg's read is that the country is now increasingly divided between a narrow group of technology clusters and a much larger domestic market still living with the consequences of the property downturn. Consumption growth has lagged GDP since the pandemic, in part because so much household wealth is tied up in real estate that is no longer appreciating. Chip fabs and data centers do not fix that balance sheet.

Nor is the AI boom big enough on its own to offset the drag. Nomura estimates the AI economy will contribute only 0.3 percentage points to Chinese GDP in 2026, a useful number to hold in mind when reading triumphalist coverage of the country's chip and model progress. Nomura's own economists have separately described a 'K-shaped divergence,' with the four top hubs, Beijing, Shanghai, Hangzhou and Shenzhen, siphoning jobs and revenue from lower-tier cities.

The honest caveat is that a single half-year is not a trend, and the piece leans on one bank's estimates for its most quotable figures. What the reporting does not resolve is whether Beijing has the fiscal or political tools to redistribute AI gains toward households before the two economies drift further apart. The upside case is that the Hefei model, patient state capital chasing a specific industry, gets copied. The downside case is that the copies do not work, and the boomtown map stays this small.