Nvidia has confirmed it is preparing to restart sales of its H20 AI chips to China after months of uncertainty caused by changing U.S. export regulations.
The company stated that it is now filing the necessary applications and expects to receive government licenses soon. Deliveries are scheduled to begin shortly after approval.
In addition, Nvidia has introduced a new chip, the RTX Pro, built specifically for the Chinese market.
According to the company, this chip fully complies with U.S. regulations and is suited for industrial applications such as smart factories and logistics.
H20
The H20 is not Nvidia’s most advanced chip; however, it is currently the most powerful product the company is allowed to sell to China under U.S. law.
It is designed for inference tasks: running existing AI models rather than training new ones. Despite its limitations, the H20 has attracted strong demand.
Major Chinese firms, including ByteDance, Alibaba, and Tencent, purchased large quantities earlier this year.
Their interest lies in the chip’s high memory bandwidth and seamless compatibility with Nvidia’s software ecosystem.
These features make it easier for developers to deploy AI models quickly and effectively.
The Ban
In April 2025, the U.S. government imposed restrictions on the sale of the H20. Officials cited national security concerns, pointing to the chip’s high memory and data transfer speeds.
The ban affected chips exceeding performance thresholds, including 1,400 gigabytes per second of memory bandwidth and 1,100 GB per second of input/output bandwidth.
Industry analysts estimated that the restriction could cost Nvidia between $15 billion and $16 billion.
Sales in China had surged earlier that year as companies anticipated such measures. However, the ban was short-lived.
Shortly after Nvidia CEO Jensen Huang attended a high-level dinner with former President Donald Trump, the administration paused the restrictions.
According to NPR, the change came after Nvidia’s pledge to increase U.S. investments in AI infrastructure.
Within days, the company announced plans to build AI data centers in the United States. The total value of the project could reach $500 billion over four years, and Nvidia will partner with global firms, including TSMC, to carry out the plan.
A China-Specific Chip
As part of its updated strategy, Nvidia unveiled a new chip tailored for the Chinese market.
The RTX Pro meets current U.S. export guidelines and targets manufacturing applications.
It is designed to support automation, robotics, and logistics: industries that are central to China’s digital economy.
This reflects Nvidia’s effort to remain compliant while continuing to serve its international customers.
By offering a market-specific solution, the company reduces regulatory risk and maintains a presence in one of the world’s largest AI markets.
DeepSeek
The debate around AI exports intensified earlier this year when DeepSeek, a Chinese startup, developed a powerful AI model using Nvidia’s older H800 chips.
These chips were banned in October 2023, yet Chinese suppliers found workarounds that allowed limited access.
DeepSeek’s achievement highlighted the limitations of current U.S. policies. Lawmakers argued that inconsistent enforcement allows Chinese firms to continue advancing AI capabilities.
This, they warned, could weaken U.S. efforts to protect critical technology.
Security and Economic Interests
Nvidia’s return to the Chinese market could pose an issue for U.S. policymakers. On one hand, the government seeks to limit the spread of advanced AI tools.
On the other hand, companies depend on global sales to fuel growth and fund innovation. The H20 situation illustrates this conflict.
National security concerns must be weighed against the economic impact on American firms.
Despite this, Nvidia remains optimistic. In a statement to TechCrunch, company spokesman Hector Marinez said CEO Huang continues to meet with officials in both Washington and Beijing.
