[Source / Observer Network Wang Yi] The Financial Times reported on May 30 that as the United States continues to tighten its export of advanced chips to China, large Chinese technology companies have begun preparing for a future without NVIDIA chips from American chip manufacturers. They are actively turning to domestic chips to develop artificial intelligence (AI).

Building on previous chip export control measures by the Biden administration, the Trump administration further tightened exports to China in April, requiring NVIDIA to obtain approval from the US government before selling its customized H20 chips for the Chinese market.

The Financial Times said that new chip orders usually take 3 to 6 months to ship, and existing NVIDIA processor inventory can only last until early next year. This has prompted large Chinese tech companies to expedite their contingency plans.

Shen Diao, executive vice president of Baidu Group and president of Baidu Intelligent Cloud Business Group, said last week that the company could choose from a range of chip options to replace NVIDIA chips, especially inference chips.

"We believe that over time, self-developed chips and increasingly efficient self-developed software stacks will together lay a solid foundation for long-term innovation in China's AI ecosystem," Shen Diao said during Baidu's first quarter 2025 earnings conference call.

Tony Liu, president of Tencent, revealed at the company's first quarter 2025 earnings conference held on the 14th that while Tencent is working to improve chip efficiency, it is also considering alternative products. He mentioned that they realized they must abandon the so-called "scale law" of American tech companies, which states that good training results can be achieved with smaller clusters. "Our current high-end chip inventory is sufficient to support the training of several generations of models."

Liu Chiping added that domestically available chips and importable chips can meet growing inference needs in some cases, rather than relying solely on purchasing GPUs (graphics processing units).

Bu Jiwen, CEO of Alibaba, also stated in his earnings conference call earlier this month that they are actively exploring diverse solutions to meet growing customer demand. While developing their own AI large models, Alibaba, Baidu, and other companies have been researching their own AI processors.

The China Modern International Relations Institute also posted in May that although US export controls are painful, they have instead triggered a wave of independent innovation in advanced AI chips in China. Huawei's Ascend chip is a prime example, with Chinese companies and institutions beginning to mass purchase and use Ascend chips.

Even if the US recently banned the use of Huawei Ascend chips "anywhere in the world," The Financial Times predicts that more Chinese tech companies will compete for this Ascend chip. Meanwhile, chips from domestic chip manufacturers such as HaiLight Information and Cambricon are currently being tested by major Chinese tech giants.

Huawei Ascend Chip - Visual China

However, the report said that transitioning systems from NVIDIA chips to domestic chips by Chinese tech companies would involve significant time and human costs. This has led most Chinese tech companies to consider a hybrid approach, where AI training continues on existing NVIDIA chips, while domestic processors are used for inference.

Facing the threat of losing the Chinese market share, American chip manufacturers are anxious. TechCrunch, an American tech website, reported on the 29th that NVIDIA and AMD are preparing simplified versions of AI chips for the Chinese market that comply with Trump administration export control regulations, which may go on sale in July.

According to reports, NVIDIA's new chip is codenamed "B20", based on its advanced Blackwell architecture but lacks high bandwidth memory (HBM), a key component for quickly processing large amounts of data. AMD hopes to create an AI accelerator based on Radeon AI PRO R9700 for sale in China.

However, The Financial Times pointed out that these new chips must not only comply with stricter export regulations set by the Trump administration but also perform better than China's domestic AI chip competitors to have a market. It remains unclear whether NVIDIA can strike a balance between the two.

Jensen Huang, CEO of NVIDIA, admitted on the 28th that when considering new products for the Chinese market, they have limited choices and cannot repeatedly reduce product performance. "We don't have specific plans yet," and even if there were plans, they would need approval from the US government.

Huang told Bloomberg on the 28th that US chip export controls to China effectively exclude NVIDIA from the Chinese market, which cannot stop China's development of AI. Instead, it will push Chinese tech companies to seek local alternatives, allowing Chinese chip manufacturers to fill the void left by American companies.

"China's competitors have evolved," Huang said, noting that the gap between American products and Chinese alternatives is narrowing, with annual increases in capacity and output from Chinese manufacturers.

As Huang said in an interview on the 19th, the idea of preventing the spread of AI technology to stop others from accessing American technology is fundamentally wrong. If American companies do not compete in the Chinese market, then Chinese technology will spread worldwide.

This article is an exclusive contribution from the Observer Network and cannot be reprinted without permission.

Original source: https://www.toutiao.com/article/7510114973603873332/

Disclaimer: The views expressed in this article are those of the author alone, and you are welcome to express your opinions by clicking the "Agree/Disagree" buttons below.