Image Source: Visual China
In the early hours of August 6, NVIDIA published a lengthy statement on its Chinese website and other channels, stating that there are no backdoors, kill switches, or monitoring software in NVIDIA chips. “These are certainly not ways to build a trustworthy system, and they never will be.”
NVIDIA indicated in the statement that to reduce the risk of misuse, some experts and policymakers have suggested the need for a “kill switch” or built-in controls in hardware to remotely disable GPUs without the user’s knowledge or consent. There are suspicions that such situations may already exist. NVIDIA emphasized that there are no kill switches or backdoors in NVIDIA GPUs, nor should there be.
According to reports from the Cyberspace Administration of China, NVIDIA’s computing chips have recently been exposed to serious security issues. Previously, U.S. lawmakers called for advanced chips exported from the U.S. to be equipped with “tracking and positioning” capabilities. Experts in the U.S. AI field revealed that NVIDIA’s computing chips have mature “tracking and positioning” and “remote shutdown” technologies. To maintain the cybersecurity and data security of Chinese users, the National Internet Information Office interviewed NVIDIA on July 31, 2025, requiring the company to explain the security risks of backdoors in the H20 computing chips sold to China and submit relevant proof materials.
Later that night, in response to this issue, NVIDIA stated: “Cybersecurity is crucial to us. NVIDIA’s chips do not have ‘backdoors’ and do not allow anyone remote access or control over these chips.”
NVIDIA’s founder and CEO Jensen Huang visited China in July to participate in events such as the Chain Expo, marking his third visit to China this year. On July 15, Interface News learned that during Huang’s visit to China, NVIDIA was submitting an application to resell the NVIDIA H20 GPU, and the U.S. government had assured NVIDIA that it would grant the license, with NVIDIA hoping to initiate deliveries as soon as possible.
Additionally, Huang announced the launch of a new and fully compatible NVIDIA RTX PRO GPU, which is “an ideal choice for creating digital twin AI for smart factories and logistics.”
The U.S. government decided in April this year to prohibit NVIDIA from selling its H20 chips to the Chinese market. The H20 was introduced to comply with U.S. export restrictions and is designed specifically for the Chinese market as an AI accelerator. Based on NVIDIA’s Hopper architecture, the H20 features advanced CoWoS packaging technology. The H20 is more suitable for vertical model training and inference but cannot meet the needs of trillion-scale model training, with overall performance slightly higher than the 910B.
During his visit to China, Huang stated in a media interview: “The Chinese market is vast, vibrant, and highly innovative, and it is also a gathering place for many AI researchers. Therefore, it is indeed crucial for American companies to establish a presence in the Chinese market.”
Unauthorized reproduction is prohibited