Huawei backs DeepSeek V4 with Ascend AI chips
The AMW Read
Novelty 2: updates DeepSeek's compute strategy and Ascend's market traction; significance 2: has segment-level impact on foundation model infrastructure and cross-segment compute supply dynamics.
Huawei backs DeepSeek V4 with Ascend AI chips
Huawei has partnered with DeepSeek to support the training and inference of its next-generation foundation model, DeepSeek V4, using Huawei's Ascend AI chips. The collaboration marks a significant integration of Chinese AI software with domestic hardware, positioning Ascend as an alternative to Nvidia's GPUs for leading Chinese AI labs.
This move reinforces the hyperscaler-distribution pattern within China's AI ecosystem, where domestic chipmakers increasingly tie their silicon to frontier model development under geopolitical export restrictions. DeepSeek gains access to a dedicated compute supply chain outside Nvidia's ecosystem, while Huawei secures a high-profile customer for its Ascend line, potentially accelerating adoption across China's AI sector.
The partnership updates the ongoing debate about the viability of non-Nvidia compute stacks for frontier model training. If DeepSeek V4 achieves competitive performance on Ascend hardware, it could validate a Chinese domestic stack and reshape the global compute supply dynamics, particularly under US export controls. However, the technical integration challenges and performance gaps remain open questions.
