Press "Enter" to skip to content

US export controls challenge Chinese tech giants’ AI goals

The U.S.-China tech rivalry intensified this week after the Biden administration updated its regulations on AI chip exports. Washington’s latest move could significantly impact Chinese tech behemoths, potentially hindering their rapid advancements in AI.

On Tuesday, the U.S. commerce department expanded the stringent export controls introduced in October 2022. The move aims to keep pace with technological progress and close any loopholes companies might exploit. Commerce Secretary Gina Raimondo emphasized that the update seeks to limit China’s access to advanced chips crucial for its military and AI breakthroughs, as per the Financial Times. This development precedes a potential summit between President Biden and President Xi, further straining U.S.-China relations.

Impact on the tech industry

Nvidia, AMD, and Intel — the three key players in the AI chip market — are feeling the heat. The high-end chips the produce are vital for training expansive AI models, leading to a surge in demand. The new regulations, however, will compel Nvidia to cease shipments to China of two processors specifically designed to adhere to the prior export controls.

Last year, in response to the initial rules, Nvidia crafted new performance-limited versions of its H100 and A100 GPUs, specifically for Chinese clientele. The U.S. is now restricting exports of these modified H800 and A800 GPUs to China under the updated regulations. Intel’s Gaudi2 chip is also included in the updated controls.

With Nvidia deriving nearly a quarter of its data center chip revenue from China, the company’s shares dipped by 6% following the announcement. Both AMD and Intel also experienced a decline.

What tightened export controls means for AI development in China

Due to these restrictions, Chinese tech groups will likely be forced to use older AI chipsets, according to the Financial Times, such as Nvidia’s V100 from 2017, to train and operate generative AI models. This poses a significant challenge, especially considering the advancements in chip technology since then, which have enabled innovations like ChatGPT.

While major Chinese tech companies have stockpiled AI chips, the long-term impact of these controls could make AI model training in China more costly and time-consuming compared to the U.S., The FT reports. Chinese tech firms such as Alibaba, Baidu, ByteDance, and Tencent had placed $5 billion-worth of orders for Nvidia chips. However, delivery delays and the new regulations might disrupt these orders.

Many believe China will still acquire some banned chips through unofficial channels, though likely not enough to meet demand.

The broader implications

The 2022 controls already had Nvidia CEO Jensen Huang concerned about the company’s constrained ability to sell its top-tier chips to China. Further restrictions could jeopardize U.S. chipmakers’ investment capabilities. Nevertheless, Huang remained optimistic, telling The FT that global demand for their products will likely offset any short-term financial impacts.

Yet, Nvidia’s near-term financial risks aren’t the only challenge, as major U.S. firms are making strategic moves into semiconductor technology. Notably, Microsoft and Google have signaled pivots to AI chip development, a long-term plan to lessening their reliance on Nvidia. Meanwhile, Intel is expanding its foundry services, though significant client announcements remain pending.

Source: ReadWriteWeb