Nvidia readies a new export-compliant chip for China but disputes the 'LPU' label

Nvidia is reportedly preparing a new chip designed to comply with U.S. export controls that it plans to ship to China in small batches by the end of 2026. The processor is built to work alongside chips already available in the Chinese market, and customers have reportedly begun placing orders, according to two Nvidia employees cited in the reporting.

The plan, if it holds, marks another attempt by Nvidia to hold ground in a market Washington’s export rules have steadily walled off. Some accounts describe the chip as a language-processing-unit, or LPU, class part aimed at inference — running trained AI models rather than training them.

Nvidia disputes that characterization. “We have no LPU sales in the China market today, and no China-specific LPU product in our roadmap,” the company said, without detailing what the reported chip is.

The stakes are set by how far Nvidia has already retreated. Chief executive Jensen Huang said in May 2026 that the company had “largely conceded” China’s AI market to Huawei, whose domestic accelerators — along with chips from firms such as Baidu — increasingly fill the gap left by restricted U.S. parts. A China-compatible inference chip had been in the works: Nvidia was reported in March 2026 to be readying such a variant to compete for that business.

Much stays unconfirmed. Nvidia has not announced the chip, its specifications or a ship date, and its denial of the LPU framing leaves the product’s actual design unclear. Whether small-batch shipments arrive by year-end will be the first test of the plan.