Deepseek announced plans to deploy at least 160,000 of Huawei's Ascend-950DT chips in a data center in Inner Mongolia. The cluster would be used for inference tasks, not training, according to Bloomberg. This marks a significant step in China's efforts to reduce reliance on Nvidia hardware.
The project would be the largest known Huawei chip cluster if completed, highlighting the company's growing role in the global AI landscape. However, Huawei faces production constraints and memory shortages, which may delay full delivery for over a year. The memory bottleneck could ease as China's top memory maker, CXMT, begins producing small batches of HBM3E.
CXMT is still three to five years behind global leaders like Samsung, SK Hynix, and Micron, which are already mass-producing HBM4. Deepseek's order is part of a broader Chinese government initiative to strengthen the domestic chip industry while maintaining AI competitiveness.
"The cluster would be the largest known Huawei chip cluster and a real step toward weaning China off Nvidia," said Bloomberg. The project underscores the strategic importance of Huawei's chip technology in the global AI race.
The announcement follows a larger push by the Chinese government to grow its own chip industry without falling behind in AI. Deepseek did not specify the timeline for full delivery, and the memory bottleneck remains a key challenge. The cluster's completion could mark a major milestone in China's tech independence.
Source: thedecoder
