Washington’s latest export restrictions on Nvidia’s H20 chips are likely to accelerate China’s shift toward domestic alternatives, as homegrown firms strive to close the gap with global rivals.

Years of controls on U.S. chips have pushed the Chinese government to invest billions in its domestic supply chain for semiconductors that are key to the development of artificial intelligence. Click to continue

Here is another comparative piece curated by Perplexity.ai:

China’s semiconductor industry is at a pivotal moment, driven by government policy, massive investment, and the urgent need to reduce reliance on foreign technology. While Chinese chipmakers are making significant progress, especially in chip design and memory, they still lag behind global leaders like Nvidia and TSMC in the most advanced manufacturing processes and AI compute power. Here’s a deeper look at how China’s chip sector compares, and the dynamics shaping its rapid evolution.

How Chinese Chipmakers Compare to Nvidia and TSMC

Government Support and Market Growth

  • China’s government has prioritized semiconductor self-sufficiency, pouring billions into the sector through initiatives like “Made in China 2025” and state-backed investment funds112. The market is expected to grow from $182.8 billion in 2024 to nearly $430 billion by 2033, with strong demand from AI, 5G, and electric vehicles11.

  • Subsidies and policy incentives are accelerating expansion in chip design, fabrication, and equipment manufacturing, with a particular focus on mature process nodes (28nm and above)312.

Progress in Chip Design and Memory

  • Chinese companies such as SMIC, Hua Hong, YMTC (NAND flash), and CXMT (DRAM) have expanded capacity and improved technology, especially in logic and memory chips12. YMTC and CXMT are now seen as strategic players in China’s push for self-sufficiency in memory storage12.

  • In chip design, firms like Huawei (with its Ascend series) and Cambricon are gaining ground, providing alternatives to Nvidia’s AI chips for domestic customers8. Cambricon’s stock, for example, has soared, reflecting investor optimism despite its much smaller scale compared to Nvidia8.

Lag in Advanced Manufacturing and Equipment

  • The most advanced AI chips, such as Nvidia’s H100 and B100, are produced using TSMC’s cutting-edge 4nm and 3nm processes-technology currently out of reach for Chinese fabs5. Chinese foundries like SMIC are competitive at mature nodes (28nm and above) but are not yet able to mass-produce chips at the leading edge12.

  • China is making progress in domestic semiconductor equipment, aiming for 50% self-sufficiency by 2025, but still depends on US, Japanese, and European suppliers for the most advanced tools42.

Impact of US Export Controls

  • US restrictions have blocked Nvidia from selling its most advanced chips (H100, B100) to China, and even the less powerful H20 chip now requires a special export license1710. In response, Nvidia is developing downgraded versions for the Chinese market, with reduced memory and performance1.

  • These controls have forced Chinese AI chip startups to design lower-power processors to retain access to TSMC and other international foundries, but the most advanced manufacturing remains largely inaccessible due to global restrictions6.

  • Despite these hurdles, Chinese tech giants like Tencent, Alibaba, and ByteDance have placed significant orders for available Nvidia chips, while also accelerating adoption of domestic alternatives110.

Investment and Future Outlook

  • China is set to remain the world’s largest investor in new chipmaking equipment in 2025, despite a projected drop in spending9. This investment is expected to further boost domestic capacity and innovation.

  • The rapid expansion in mature node manufacturing is disrupting the global supply-demand balance and prompting investigations and potential trade actions from the US and its allies12.

  • Sustainability and energy efficiency are becoming central to China’s semiconductor strategy, with firms adopting greener manufacturing practices to align with global standards and attract investment11.

Head-to-Head Comparison: China vs. Global Leaders

Feature/CapabilityChina (SMIC, Huawei, YMTC, etc.)Nvidia/TSMC (Global Leaders)
Process Node28nm and above (mature nodes), limited 7nm pilot production4nm (Nvidia H100), 3nm (B100/B200)
AI Chip PowerImproving (Huawei Ascend 910B, Cambricon), but lower compute and efficiencyIndustry-leading (Nvidia H100/B100)
Memory TechnologyYMTC (NAND), CXMT (DRAM) catching upLeading-edge HBM and memory solutions
Equipment Self-SufficiencyTargeting 50% by 2025, but still reliant on foreign techFull access to global advanced tools
Government SupportMassive subsidies, policy focusMarket-driven, some US support
Market Share (China)Growing rapidly, especially in domestic marketNvidia had 90% of China’s AI chip market pre-controls6

Key Takeaways

  • China is closing the gap in chip design and memory, and is investing heavily in mature-node manufacturing, but remains behind in the most advanced chipmaking and AI compute power512.

  • US export controls are accelerating China’s push for self-reliance, but also limit access to critical tools and processes, slowing the pace of technological catch-up1710.

  • Domestic alternatives are improving, with companies like Huawei and Cambricon gaining traction, but they are not yet a full substitute for Nvidia’s most advanced chips810.

  • The race is dynamic: As China ramps up investment and innovation, the global semiconductor landscape is becoming more fragmented, with parallel supply chains and growing competition911.

China’s chipmakers are catching up fast, but the technological and geopolitical barriers ensure that the race with Nvidia and TSMC will remain fierce and unpredictable for years to come.