Nvidia announced on Monday that it has applied for licenses to resume selling its H20 artificial intelligence chips to China.
This decision comes after months of uncertainty caused by U.S. export restrictions that were introduced and then unexpectedly eased following high-level meetings between Nvidia and U.S. officials.
According to a company blog post, Nvidia expects to receive government approval soon and begin shipments shortly after. In addition to restarting H20 sales, Nvidia is launching a new chip called the RTX Pro, specifically designed for China. The company says this chip fully complies with export rules and is well-suited for industrial uses such as smart factories and logistics systems.
The H20 chip has become central to the growing tech tension between the U.S. and China. Although it’s not Nvidia’s most advanced processor, it is currently the most powerful AI chip the company is allowed to sell in China under current U.S. restrictions. The H20 is designed mainly for inference tasks — where AI models are used in real-world applications — rather than for training large-scale AI systems.
Earlier this year, major Chinese firms like ByteDance, Alibaba, and Tencent rushed to buy up Nvidia’s AI chips in anticipation of tighter export rules. These chips are popular not just for their performance but also because of Nvidia’s robust software ecosystem, which makes them easier to use than domestic alternatives.
The situation took a major turn in April when the Trump administration blocked sales of the H20, targeting chips that exceeded specific memory and bandwidth limits. That decision was expected to cost Nvidia as much as $16 billion in potential revenue, based on first-quarter demand from China. But the restrictions were short-lived. After Nvidia CEO Jensen Huang attended a $1 million-per-guest dinner at Trump’s Mar-a-Lago resort, the administration reversed course. NPR reported that the decision to pause the ban came after Nvidia pledged to invest heavily in U.S.-based data centers.
Following the policy change, Nvidia announced plans to build AI server infrastructure in the U.S. worth up to $500 billion over four years, working with partners such as chipmaker TSMC. The sudden shift has drawn criticism from U.S. lawmakers, who say the back-and-forth weakens efforts to limit China’s access to advanced AI technology. They point to Chinese company DeepSeek, which recently developed a strong AI model using Nvidia’s older H800 chips, as evidence of how quickly China can benefit from high-end hardware.
In a statement to TechCrunch, Nvidia spokesperson Hector Marinez said that CEO Huang has been in talks with officials in both Washington and Beijing, highlighting the positive role AI can play in business and society.
If this article helped you, please consider supporting our work. Every small contribution keeps Abijita.com independent and running.
This episode reflects the ongoing struggle in U.S. tech policy: balancing national security with the economic interests of powerful companies. Given the events of 2025 so far, it’s likely we’ll see more policy reversals like this in the future.
