Back to News
Market Impact: 0.7

Home-grown heroes: how Huawei and DeepSeek are helping China break reliance on US chips

NVDABABATSM
Artificial IntelligenceTechnology & InnovationSanctions & Export ControlsTrade Policy & Supply ChainGeopolitics & WarCompany FundamentalsCorporate EarningsProduct Launches

China is intensifying its drive for AI chip self-sufficiency, with DeepSeek pivoting to domestic hardware and Huawei unveiling an ambitious roadmap for its Ascend AI chips and supercomputing clusters designed to rival Nvidia's offerings. This strategic shift, spurred by US export restrictions, has already impacted Nvidia's stock and underscores China's commitment to reducing reliance on foreign technology, despite ongoing challenges in domestic production capacity and single-chip performance compared to leading US counterparts.

Analysis

China is aggressively accelerating its AI hardware self-sufficiency, a strategic shift catalyzed by US trade restrictions. This is demonstrated by AI start-up DeepSeek's pivot towards domestically produced chips and, more significantly, by Huawei's detailed roadmap for its Ascend AI chip series. This development has already registered as a headwind for Nvidia (NVDA), whose shares declined on the news. Huawei's strategy aims to circumvent US sanctions by doubling the compute power of its Ascend chips annually and creating powerful supercomputing clusters, like the Atlas SuperPods, which can aggregate up to a million proprietary NPUs. While Huawei's single-chip performance is acknowledged to be "one or two generations behind" Nvidia, the company plans to mitigate this gap through clustering and chip stacking. The collaboration is already tangible, with Huawei's Ascend chips being used for inference in DeepSeek's R1 model. This domestic ecosystem is gaining momentum, evidenced by Cambricon Technologies' reported 4,348% year-over-year revenue increase and efforts from tech giants like Tencent and Alibaba (BABA) to adapt their infrastructure to domestic hardware. However, significant execution risks remain, primarily concerning the production capacity and yield of domestic foundries like SMIC, which are tasked with manufacturing these advanced 7-nanometre chips at scale, a feat that will be challenging to achieve at levels comparable to industry leader TSMC.

AllMind AI Terminal

AI-powered research, real-time alerts, and portfolio analytics for institutional investors.

Request a Demo