DeepSeek, an AI lab based in Hangzhou, plans to transition its hardware infrastructure to Huawei processors. According to CEO Liang Wenfeng, Huawei will begin supplying AI training chips to the company between Q4 2026 and Q1 2027.
The lab is considering the deployment of at least 160 000 Ascend 950DT chips at a new data center in Ulanqab, Inner Mongolia. This facility is expected to require approximately 1 gigawatt of power to operate.
DeepSeek previously tested Huawei hardware in June 2026, when it used 1 000 Ascend 910C chips to complete post-training for its V4 Pro model. The company also collaborates with Huawei on software optimizations, including CANN adaptations.
Huawei has accelerated the release of its next-generation Ascend 960DT training chip to Q1 2027, moving the launch date three quarters ahead of the original schedule to meet rising demand from domestic AI developers.
