Nvidia’s AI accelerator shipments in China fell 44 percent year over year in the first half of 2026, according to IDC data reported by Reuters and summarised by Semiconductor Engineering. The decline leaves Nvidia with 34 percent of China’s data centre AI accelerator market by units shipped, against 28 percent for Huawei, a gap that would have been unthinkable two years ago.
The numbers are the shipment level evidence for what export controls were designed, and not designed, to do. Controls have cut Nvidia’s volume sharply, but the demand did not disappear with it. It migrated, principally to Huawei, whose share now sits within six points of the market leader in the world’s largest national AI market outside the United States.
The same industry review notes the wider contest tightening around Nvidia. Chinese memory maker CXMT reportedly plans a DDR5 server module using a new cell design by year end, and a Chinese prototype EUV light source has been unveiled with mostly localised components, although it remains far below commercial power levels and years from mass production. None of that replaces Nvidia’s stack today. Together it shows the direction of travel: a parallel supply chain being built under pressure, one shipment share point at a time.