DeepSeek’s advancements in artificial intelligence (AI) models are giving Chinese semiconductor companies, including Huawei, a stronger foothold in the domestic market, allowing them to compete more effectively against powerful U.S. processors.
For years, Huawei and other Chinese firms have struggled to match Nvidia in developing high-end chips capable of handling AI training, a process that involves feeding data to algorithms to refine decision-making capabilities. However, DeepSeek’s models emphasize “inference,” the phase where AI generates insights, optimizing computational efficiency instead of relying purely on processing power.
Read more: BYD Unveils DeepSeek AI-Powered Self-Driving Tech, Aiming to Rival Tesla in Budget EVs
Analysts suggest that this focus on inference could help bridge the performance gap between Chinese AI processors and their superior U.S. counterparts. In recent weeks, several Chinese AI chipmakers, including Hygon, Tencent-backed EnFlame, Tsingmicro, and Moore Threads, have announced support for DeepSeek models, though specific details remain scarce. Huawei declined to comment, and other firms did not respond to media inquiries.
Industry experts predict that DeepSeek’s open-source nature and affordability could accelerate AI adoption and real-world applications, helping Chinese firms navigate U.S. export restrictions on advanced chips. Even before DeepSeek’s recent prominence, Huawei’s Ascend 910B chip had already been recognized as more suitable for inference tasks, such as those used in chatbots, rather than intensive AI training.
Numerous Chinese companies across various industries, from automakers to telecom providers, have expressed plans to integrate DeepSeek’s models into their operations. According to Lian Jye Su, a chief analyst at Omdia, this shift aligns well with the strengths of Chinese AI chip vendors, as inference tasks require less raw computing power and more localized, industry-specific expertise.
Nvidia Maintains Market Dominance
Despite Chinese AI chips being cost-effective for inference applications within China, Nvidia remains the global leader. While U.S. export restrictions prevent the sale of Nvidia’s most advanced AI training chips to China, the company still provides less powerful versions that Chinese firms can use for inference tasks.
On Thursday, Nvidia published a blog post highlighting how inference time is increasing under new AI scaling laws, arguing that its chips remain essential for advancing DeepSeek and other reasoning models. Beyond hardware, Nvidia’s CUDA platform—designed for parallel computing and AI processing—has solidified its industry dominance.
While many Chinese AI chipmakers have claimed CUDA compatibility, Huawei has been the most aggressive in developing an alternative, Compute Architecture for Neural Networks (CANN). However, experts believe the transition away from CUDA will be challenging due to its extensive software library and long-term investment in developer support. Omdia’s Su notes that Chinese AI chip firms currently lag in software performance, further reinforcing Nvidia’s competitive edge.
