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

As China's push for technological self-reliance accelerates, more Chinese companies are moving away from dependence on U.S. chipmakers like NVIDIA, opting instead for domestic or alternative suppliers... Read original →

China’s accelerated push for AI chip self-reliance is triggering a deep-stack technical realignment: autonomous driving algorithms are migrating from CUDA to domestic architectures like Ascend and Hanguang, forcing compiler and toolchain innovation. U.S. export controls have inflated NVIDIA’s high-end chip costs by over 30%, compelling automakers like NIO and XPeng to embed in-house silicon into BOM models—supply chain security now outweighs raw performance. NVIDIA may counter with Orin price cuts or local JVs, but Huawei’s 7nm automotive chips integrated with HarmonyOS cockpit already form a defensible moat. Within 12–24 months, if China achieves 3nm pilot production before 2027 despite EUV constraints, its AI training chip market share could surge from under 5% to 25%, fundamentally redrawing the global compute geopolitics map.