Huawei has set its sights on the annual release of new AI chips in the Ascend series, with twice as much performance each generation. This was stated by the company’s vice president and head of cloud business in China, Chen Lin, noting that the industry has faced a shortage of computing resources and an increase in memory costs, which is hindering model learning. Stable supplies and a predictable schedule for the upgrade, in his view, will be the answer to these challenges.

The first step will be to launch the Ascend 950DT chip in the Huawei Cloud service in August 2026. The company promises a significant increase in vector performance, memory capacity and interconnection speeds, as well as native support of modern formats such as FP8 - this will simplify the development and fine-tuning of models, including autonomous driving scenarios.

In parallel, Huawei Cloud announced the creation of a global computing and storage network based on its own Ascend chips and localized supply chain. Plans cover 34 regions and 102 accessibility zones; large clusters are already operating in Guiana, Wuhu, and Inner Mongolia. The company is also strengthening its focus on automotive and corporate digitalization, offering "smart" car manufacturers a safer cloud infrastructure. Given these claims, it is worth noting that as early as January, Google DeepMind chief Demis Hassabis admitted: China has almost caught up with the West in artificial intelligence.