China’s AI Labs Unveil ASIC Designs to Offset Nvidia GPU Restrictions
NVDA•China's top AI developers, including Z.ai under CEO Zhang Peng, are designing custom ASIC inference chips to counter tighter US controls on Nvidia GPUs. Huawei's homegrown alternatives are ramping production but still trail Nvidia's GPUs in throughput and energy efficiency.
1. Custom ASIC Chip Initiatives
Several leading Chinese AI model developers, led by Z.ai under CEO Zhang Peng, have launched in-house chip divisions focused on bespoke ASIC designs optimized for inference workloads. This shift aims to improve power efficiency and unit economics as demand for AI services grows.
2. US Export Controls Tighten GPU Access
Since late 2023, the US has restricted exports of high-end Nvidia GPUs, including A100 and H100 series, curbing Chinese companies’ ability to procure these accelerators. The resulting shortages have accelerated domestic chip projects.
3. Huawei’s Domestic AI Chip Progress
Huawei has rolled out its own AI accelerators to fill the GPU gap, securing initial production volumes for local data centers. Early comparisons indicate these chips lag Nvidia devices in raw throughput and energy efficiency, slowing customer migration.
4. Implications for Nvidia’s China Business
If Chinese labs and cloud providers scale up custom ASIC adoption, Nvidia could see reduced GPU sales growth in its second-largest market. Sustained domestic innovation may reshape the competitive landscape over the next two years.
