Chinese Regulatory Action Restricts ByteDance's Access to Nvidia Chips: Implications for AI Development and Technological Competition

Recent regulatory actions in China are significantly impacting ByteDance, the parent company of TikTok and Douyin, by restricting its access to Nvidia's advanced AI chips. This development is multifaceted, stemming from both U.S. export controls and China's push for technological self-sufficiency, and carries considerable implications for AI development and the broader landscape of technological competition.

Background: U.S. Export Controls and China's Response

Since 2022, the U.S. government has implemented increasingly stringent export controls on advanced semiconductors, particularly Nvidia's high-performance GPUs, to China. These restrictions aim to curtail China's access to cutting-edge technology that could be used for military modernization and AI development. Nvidia, in an attempt to maintain its market presence in China, designed modified chips like the H800 and H20 that complied with U.S. regulations. However, even these "watered-down" chips have faced increased scrutiny and restrictions.

In response to these constraints, China has ramped up its efforts to develop a domestic semiconductor industry and reduce reliance on foreign technology. The Chinese government has provided substantial subsidies and support to domestic chip manufacturers like Huawei and Cambricon, encouraging them to produce competitive AI chips.

The Regulatory Action and Its Impact on ByteDance

Recent reports indicate that Chinese regulators have now prohibited ByteDance from utilizing Nvidia's GPUs in its new data centers. This decision aligns with an August directive instructing local firms to halt new orders of Nvidia AI chips and prioritize domestic alternatives. The Cyberspace Administration of China (CAC) has reportedly banned the purchase of Nvidia's AI chips, including the RTX Pro 6000D, a chip specifically designed for the Chinese market, by major technology companies like ByteDance and Alibaba. Furthermore, new data center projects receiving state funding are now required to use domestically-made AI chips.

This regulatory action presents several challenges for ByteDance. First, it leaves the company with a potentially large inventory of Nvidia chips that it cannot use in its new data centers. ByteDance was reportedly the largest Chinese buyer of Nvidia chips in 2025, anticipating potential U.S. restrictions. Second, ByteDance must now transition to using domestic AI processors, which may not currently match the performance of Nvidia's high-end GPUs. This shift could impact the efficiency and capabilities of ByteDance's AI services, including TikTok, Douyin, and its expanding AI applications.

Circumventing Restrictions: Offshore Training and Domestic Innovation

Faced with these restrictions, Chinese tech giants like ByteDance and Alibaba have increasingly turned to training their AI models in data centers located in Southeast Asia. By leasing computing power from foreign-owned data centers in countries like Singapore and Malaysia, these companies can access high-end Nvidia GPUs while remaining compliant with U.S. export controls. This workaround allows them to continue developing advanced AI models with performance targets comparable to those of Western AI labs. The trained models can then be deployed within China using domestically sourced chips for inference.

In addition to offshore training, ByteDance and other Chinese companies are also investing heavily in domestic chip development. DeepSeek, for example, is collaborating with Huawei to optimize AI training using local silicon. Huawei's Ascend 910C chip reportedly delivers approximately 60% of the performance of Nvidia's H100 AI chip, and Cambricon's MLU 590 achieves 80% of the performance of Nvidia's A100. While these domestic chips may not yet match Nvidia's top-tier offerings, they are rapidly improving, and Chinese companies are finding ways to compensate for lower individual performance through scale.

Implications for AI Development and Technological Competition

The Chinese regulatory action and the resulting adaptations have significant implications for AI development and technological competition:

• Accelerated domestic innovation: The restrictions on Nvidia chips are incentivizing Chinese companies to accelerate the development and adoption of domestic AI chips. This push for self-sufficiency could lead to breakthroughs in chip design and manufacturing, strengthening China's position in the global semiconductor industry.
• Geopolitical shift in AI infrastructure: The rise of a parallel AI ecosystem in China, powered by domestic chips, could lead to a bifurcation of global AI infrastructure. This split could have long-term consequences for technology standards, data governance, and the overall direction of AI development.
• Continued demand for Nvidia chips: Despite the restrictions, Chinese companies continue to seek access to Nvidia's advanced GPUs through offshore training and other means. This sustained demand highlights Nvidia's dominance in the AI training market and suggests that the company will continue to find ways to serve the Chinese market, albeit through indirect channels.
• Impact on technological competitiveness: The restrictions and the responses to them are reshaping the competitive landscape in AI. While China may face short-term challenges in accessing the most advanced chips, its long-term investments in domestic innovation could lead to a more competitive and resilient AI ecosystem.

The situation remains dynamic, with the U.S. government considering whether to allow exports of Nvidia's H200 AI chip to China. However, the broader trend suggests that China is committed to building a self-reliant AI ecosystem, even if it means navigating complex regulatory hurdles and investing heavily in domestic chip development. The long-term impact of these developments on global AI leadership and technological competition remains to be seen.