Chinese authorities have permitted ByteDance and Tencent to receive approximately 10,000 Nvidia H200 AI processors apiece over the past several weeks, according to sources familiar with the matter cited by the Financial Times. This allowance marks a calculated decision by Beijing to provide its prominent technology firms with advanced computing power necessary for training sophisticated AI models. The H200, while not Nvidia's most advanced chip, offers a substantial performance increase over previous generations available to China, such as the H20.
The United States government had previously restricted the sale of high-performance AI chips to China due to concerns over their potential military applications. However, in December 2025, the U.S. Department of Commerce's Bureau of Industry and Security (BIS) revised its licensing policy, allowing for case-by-case review of export license applications for the Nvidia H200 and similar chips to approved Chinese customers. This policy change followed an announcement by then-President Trump in December 2025. The conditions for approval included demonstrating that exports would not reduce global semiconductor production capacity for U.S. customers, that Chinese purchasers had export compliance procedures, and that products underwent third-party security testing in the U.S.
Despite the U.S. easing its export controls, China had maintained its own restrictions on the import of these chips, aiming to foster its domestic semiconductor industry. Huawei currently leads China's domestic AI chip market, but its offerings still lag Nvidia's in performance. The recent shipments of H200 chips suggest a recalibration of China's strategy, balancing the push for indigenous chip development with the immediate need for advanced foreign hardware to remain competitive in the global AI race.
Other Chinese technology companies are also expected to receive approvals for similar-sized H200 shipments. The U.S. had authorized certain Chinese firms to purchase up to 100,000 H200 units each. However, Beijing has reportedly instructed companies to deploy most of these newly acquired chips outside mainland China, specifically in Hong Kong, to continue supporting domestic chipmakers.
This directive presents a logistical challenge for Chinese tech groups, as Hong Kong currently lacks sufficient data center capacity and adequate power supply to house a large volume of these processors. One individual familiar with the situation described it as a "dilemma," noting the demand for chips but the difficulty in utilizing them within Hong Kong.
Nvidia reportedly holds an inventory of approximately 500,000 H200 chips, largely earmarked for Chinese clients, but deliveries were previously delayed due to Beijing's import restrictions. Some of Nvidia's partners, such as Lenovo, have recently informed Chinese customers that they can resume placing orders for products containing H200 chips. These partners integrate Nvidia chips with CPUs and other components to sell as AI servers.
The H200 chip is built on Nvidia's Hopper architecture and is designed for memory-intensive workloads, making it suitable for generative AI, large language models, and advanced analytics. It provides roughly six times the performance of the H20, which was previously the most powerful Nvidia chip available to Chinese customers. The H200 remains at least two generations behind Nvidia's most advanced chips, such as the Blackwell B200, which are still restricted from export to China.
This development underscores the complex interplay between national security concerns, economic competition, and technological advancement in the semiconductor industry. While China continues its long-term goal of reducing reliance on foreign semiconductors, the immediate need for advanced AI training capabilities appears to have prompted a tactical adjustment in its import policies.