DeepSeek looks to reduce reliance on Nvidia with own AI chip - Yahoo Finance

DeepSeek is developing its own AI chip to reduce reliance on Nvidia amid U.S. export restrictions, reflecting a broader trend among global tech firms seeking technological autonomy. While Nvidia's chi... Read original →

DeepSeek’s in-house AI chip initiative isn’t merely a workaround for Nvidia’s export-blocked GPUs—it signals a forced vertical integration across China’s AI stack. A 3nm design reliant on TSMC (Taiwan, China) faces acute geopolitical fragility, while domestic fabs lack the yield and performance for large-scale training. This pressure will accelerate co-evolution of EDA tools, advanced packaging, and open-source frameworks, creating new technical interdependencies. Compliance burdens now extend beyond tariffs to full-stack re-engineering—IP licensing, equipment access, even talent mobility. In response, Nvidia may flood China with further downgraded H20 variants, while Google and Microsoft tighten custom silicon licensing. Within 18 months, if DeepSeek successfully couples its chip with proprietary models, it could trigger a wave of Chinese AI firms abandoning general-purpose GPUs in favor of heterogeneous architectures paired with open models—potentially fracturing the global AI infrastructure into competing technological blocs.