The launch of DeepSeek V4 marks a significant turning point in the AI landscape, not only at the model level but across the entire underlying infrastructure. For the first time, China is approaching a fully self-sufficient AI ecosystem, operating without reliance on U.S. technology.
Demand for the Ascend 950 chip series from Huawei has surged immediately following the V4 announcement. According to industry sources, major tech players including Alibaba, Tencent, and ByteDance are actively placing new orders. Cloud providers and GPU rental platforms are also rushing to secure supply.
What stands out is that DeepSeek V4 has been specifically optimized to run entirely on domestic chips, marking the first time a large-scale Chinese AI model operates without GPUs from NVIDIA or other U.S.-based technologies. This represents a major step forward in China’s long-term strategy for technological independence.
To support this ecosystem, Huawei has developed the Ascend SuperNode infrastructure, built on Ascend 950 chips, to handle V4 inference workloads. At the same time, major cloud platforms such as Alibaba Cloud and Tencent Cloud deployed V4 almost immediately upon release, enabling access for millions of users and developers. This rapid rollout is significantly increasing AI query volumes and, in turn, driving demand for underlying hardware.
From a technology standpoint, the Ascend 950PR variant is among the few domestic chips capable of supporting compressed AI computation formats, allowing more operations per second at lower cost. While it still lags behind high-end chips like H200, it already outperforms the H20, previously the most advanced NVIDIA product available in China.
However, challenges remain. Supply of high-end domestic AI chips is still constrained, and export restrictions on advanced semiconductor manufacturing equipment continue to limit China’s ability to scale production. Huawei is expected to ship around 750,000 units of the 950PR this year, with mass production starting in April and large-scale deliveries ramping in the second half of 2026.
DeepSeek V4 is more than just a breakthrough in AI models. It is acting as a catalyst for the entire domestic hardware ecosystem in China. As chips, cloud platforms, and AI models evolve together, a new integrated structure is beginning to take shape.
This is no longer just a competition in performance. It is a competition in controlling the full AI value chain, from silicon to application.
