Huawei chip breakthroughs are fueling a resurgence in China’s chip industry, with local semiconductor firms rapidly closing the gap with global giants. In Shanghai, traders joke that Cambricon, a Chinese chip designer, competes not only with Nvidia’s processors but also its stock performance. Cambricon’s value is much smaller, yet its shares have surged 350% in a year—far exceeding Nvidia’s growth.
This momentum comes amid tightening U.S. export restrictions. In April, authorities blocked Nvidia’s H20 chip from entering China, despite it being crafted to comply with previous regulations. The Biden administration plans to cancel a complex export licensing framework, originally set for May 15. However, any replacement—likely a collection of bilateral deals—offers limited benefit to Chinese tech firms.
Still, China’s AI development shows no signs of slowing. Giants like Alibaba and Tencent continue building AI data centers at full speed. Their progress stems from two main developments. First, Chinese startup DeepSeek has optimized AI model efficiency, allowing more work with fewer chips. Second, supply networks—though unofficial—are reportedly providing Nvidia hardware despite trade bans.
Huawei’s chip breakthroughs stand at the center of this transformation. The company’s new CloudMatrix cluster connects 384 Ascend AI chips using advanced networking. Early reports say it can outperform Nvidia’s NVL72, though it consumes more power. Huawei’s Ascend 910C chip has already gained attention for its performance, despite relying partly on foreign components. Cambricon and Hygon have also launched or tested Nvidia alternatives, showing a wider shift in local AI chip design.
Beyond processors, Chinese firms are advancing in high-bandwidth memory (HBM)—essential for AI performance. Though SK Hynix, Samsung, and Micron dominate the sector, Chinese firm CXMT is quickly gaining ground. Progress in chipmaking tools is also notable. AMEC launched an etching tool for NAND memory, challenging Lam Research. Naura has developed silicon-germanium layering tech, competing with U.S. firm Applied Materials.
Challenges remain. Most local chip buyers are state-owned. Huawei’s chips still rely on global supply chains, and China lacks a domestic version of ASML’s advanced lithography machines. TSMC, the top global chipmaker, cannot supply China due to U.S. restrictions. Some reports claim Huawei accesses TSMC wafers through third-party channels—allegations both companies deny.
Software is another hurdle for Huawei chip breakthroughs. Nvidia’s CUDA remains the industry standard, used by nearly all AI developers. It only works with Nvidia hardware, and switching platforms breaks access to a global ecosystem. Huawei’s answer, CANN, supports Ascend chips but lags years behind CUDA. Developers describe it as buggy and hard to adopt.
Despite these barriers, Chinese firms are increasingly embracing local technologies. Previously, many avoided domestic chips due to high costs and poor performance. Today, political pressure and strategic urgency have changed minds. Companies now view chip independence as essential for survival. That shift is powering further investment and innovation.
With Huawei chip breakthroughs leading the charge, China’s chip industry is gaining ground faster than many expected. Though it still faces external dependencies and technical gaps, its rapid progress signals a new era of self-reliance and ambition in the global tech race.
