On July 15, 2025, NVIDIA’s CEO Jensen Huang confirmed that the U.S. government has approved export licenses for the H20 AI chips to China. This crucial development ends a period of export restrictions and allows NVIDIA to resume shipments, reinforcing its position in one of the world’s fastest-growing AI ecosystems.
“We’ve received our export license and can start shipping immediately. This is truly great news for us and for the AI community in China,” said Huang.
The H20 chip is a specialized AI accelerator built on NVIDIA’s Hopper architecture, designed specifically for the Chinese market to comply with U.S. export regulations introduced in 2023. While the H20 delivers roughly one-sixth the compute power of the flagship H100, it offers superior memory bandwidth and capacity compared to the older A100, making it ideal for local AI model deployments.
Chinese AI startup DeepSeek has leveraged the H20 chip to achieve significant cost reductions in training and inference by optimizing its algorithms. Their R1 671B model runs natively on a single FusionServer G8600 server equipped with 8 H20 GPUs, demonstrating impressive performance without compromis
March 2025: Reports of tight H20 chip supply in China
April 16, 2025: U.S. imposes formal export ban on H20; NVIDIA writes down $5.5 billion in inventory
Q1–Q2 2025: NVIDIA experiences $13.5 billion revenue loss due to export restrictions
With the export license approval, NVIDIA aims to recover its market share in China. Jensen Huang highlighted that China hosts 50% of global AI researchers and emphasized the country’s vital role in AI innovation.
Additionally, NVIDIA announced the upcoming RTX Pro GPU, targeting professional workloads in computer graphics, digital twins, and AI applications.
The lifting of export restrictions for the H20 chip represents a significant win for both NVIDIA and Chinese AI developers. It highlights the importance of balanced trade policies that foster technological collaboration while addressing regulatory concerns.