Huawei aims to deliver faster AI chips, faster

Huawei is pulling forward the release of two Ascend 960 AI chip models by three to nine months, deputy chairman David Wang said at the Huawei Connect 2026 conference. The Ascend 960DT will now ship in the first quarter of 2027, three quarters earlier than planned, and the Ascend 960R in the third quarter of 2027, three months ahead of schedule. Huawei is concentrating on a cluster-based architecture that assembles SuperPoDs of up to 4,096 optically interconnected neural processing units into SuperClusters of up to 512,000 NPUs, as it lags Nvidia in individual chip processing power and faces US trade sanctions.

Huawei is accelerating its AI chips development, bringing forward the release of the next two models in the family powering its AI computing clusters by three to nine months. Its Ascend 960 chip family is a major component of supercomputing portfolio. It now plans to release the Ascend 960DT in the first quarter of 2027, three quarters earlier than previously planned, while the Ascend 960R will be introduced in the third quarter of 2027, three months ahead of schedule, Huawei deputy chairman David Wang said at the Huawei Connect 2026 conference https://www.huawei.com/en/news/2026/9/hc-wang-keynote . The company’s is lagging behind Nvidia when it comes to the processing power of individual chips, so it is concentrating its efforts on developing an architecture based on clusters. This focus on architectural innovation will, Huawei said, enable it to build a solid computing foundation on its 960 chips by assembling SuperPoDs, each containing up to 4,096 optically interconnected neural processing units, into SuperClusters containing up to 512,000 NPUs. Outside its home market of China, Huawei is battling with US trade sanctions https://www.opensanctions.org/entities/NK-P3xHZasb3VZasvJLDwiGHC/ amid suggestions that the company has deliberately downplayed the power of its individual chips https://www.networkworld.com/article/4004178/huawei-says-it-trails-its-us-rivals-in-chips-but-is-closing-the-gap.html to avoid the restrictions. Its cluster-based architecture could be a way to create a product powerful enough to attract the attention of customers, but not so powerful it attracts that of regulators.