CNBC's The China Connection newsletter: China learns to build without Nvidia - MSN

Recent reports on China's semiconductor industry development reveal strategic responses to international technology sanctions. As US-China tech rivalry intensifies, China is striving to reduce depende... Read original →

U.S. AI chip export controls are catalyzing a fundamental restructuring of China’s semiconductor ecosystem. Technically, domestic GPU firms are bypassing CUDA by embracing RISC-V and proprietary architectures, accelerating EDA and advanced packaging innovation—yet the EUV blockade still throttles sub-3nm logic scaling. Compliance-wise, even indirect access to TSMC via Taiwan, China or third parties now carries elevated geopolitical scrutiny, inflating supply chain costs and delivery risks. Strategically, while NVIDIA retains dominance in high-end AI training, its China-specific chips like H20 yield minimal margins; mass adoption of domestic alternatives from Ascend or Cambricon in inference could reshape global AI chip demand. Over the next 12–24 months, China may achieve 'performance detours' through mature-node stacking and chiplet integration—but the true long-tail impact is bifurcation: the emergence of two parallel semiconductor standards, forcing multinationals into alignment choices and driving systemic R&D redundancy and capex inflation across the industry.