In mid-July 2025, the American semiconductor chip giant Nvidia announced the U.S. government would allow it to sell its advanced H20 artificial intelligence (AI) chip to China, following previous export controls imposed in April. Commerce Secretary Howard Lutnick explained the move, saying “you want to sell the Chinese enough that their developers get addicted to the American technology stack.” Lutnick's comments echoed what Nvidia company officials had noted earlier in the year, that it was “Better to have [Chinese] customers paying Nvidia billions of dollars and remaining hooked on its chips … than to send them searching for a Chinese alternative.”
However, China's own stated industrial policy indicated a long-term goal of weaning the semiconductor sector from American know-how, both to avoid dependence on U.S. technology and to hedge against recurring export controls. As a result, China's chip ties with the U.S. have seen many twists and turns over the past several years, and the technology competition between the two nations shows an unclear path going forward.
China's government likely sensed the U.S. desire to keep its semiconductor sector dependent on American technology as early as 2022, when the State-owned Assets Supervision and Administration Commission (SASAC) took steps to advance a policy dubbed “Delete America,” or “Delete A” for short. The SASAC Document 79 targeted foreign software in IT systems, which the document said should be replaced by 2027. But some Chinese semiconductor makers latched on to the “Delete A” idea to reduce foreign dependence within the technology supply chain. The drive to localize was summed up in the slogan “Xinchuang,” or IT Innovation.
To support the home-grown semiconductor sector, China created its National Integrated Circuit Industry Investment Fund in 2014. This so-called “Big Fund” directed support to build China's semiconductor supply chain, and was supported by the Ministry of Finance, large state-owned banks including ICBC and the China Construction Bank, and local governments in Beijing, Shanghai, and other major cities. Three rounds of funding contributed some $95 billion to the fund, with the latest in May 2024 adding $48 billion.
As of early 2024, China had begun to dominate the global market for legacy chips, defined as ones of the 28-nanometer or larger generation. These chips are currently used in products such as automobiles, medical devices, and home appliances. In contrast, computers used for faster and more complex functions, such as for AI, employ as small as 2-nanometer to 4-nanometer process nodes. By 2027, China's legacy chips were forecast to grow to 39 percent of the global market, with 2024 seeing more added production capacity than the rest of the world combined.
As China's industrial policy saw results in the legacy chip sector, the government forged ahead with its “Delete A” efforts in several key technology areas. In early 2024, the Ministry of Industry and Information Technology directed the country's largest telecom carriers to phase out use of foreign semiconductors in their networks by 2027. The biggest foreign losers would be American chipmakers Intel and Advanced Micro Devices (AMD). Official guidelines for government agencies and state-owned enterprises discouraged them from purchasing laptops and desktop computers with chips from either company. Of eight options for purchasing central processing units (CPUs), the top six were domestic Chinese CPUs, and the last two were Intel and AMD.
The results of such adverse measures caused significant financial setbacks for the two American chipmakers in the China market. Intel's revenue from China declined from $20.03 billion in 2019 to $14.85 billion in 2023, while AMD's China revenue fell from $5.21 billion in 2022 to $3.42 billion the following year. The results were significant, as China was Intel's largest market, making up 27% of the company's revenue in 2023. After AMD was restricted by the US from selling high-end AI chips to China, its revenue from sales in China fell from 22% in 2022 to 15% in 2023.
As the effects of China's industrial policy took hold, domestic chipmakers like Huawei Corporation, with their Kunpeng 920 chips, gained market share in the telecom market. The more advanced 7-nanometer Kunpeng, fabricated by Taiwan's TSMC process, gave the Chinese company access to more sophisticated chips. As Chinese mobile phone service providers moved from fifth generation (5G) to faster network protocols, in October 2023 the major operating company China Telecom bought 4000 AI servers, with 53% powered by Intel's CPUs, but the rest by Huawei's Kunpeng processors. Overall, Huawei reported a doubling of operating profit from 42.2 billion yuan (roughly $6.3 billion) in 2022 to 104.4 billion ($14.8 billion) in 2023, though 2024 saw a decline to 62.6 billion yuan ($8.6 billion).
China's chip industrial policy has not seen all rosy results. One report in 2022 found some $2.3 billion in semiconductor investments, much of it government funding, went to projects that never produced a single chip. From 2019 to 2022, six new major chip building projects, including Wuhan Hongxin Semiconductor Manufacturing Corporation and Quanxin Integrated Circuit Manufacturing in Jinan, failed.
Despite China's dominance in the legacy chip sector, the nation depended on imports for most of its advanced chip requirements. As of early 2024, Chinese cloud companies sourced some 80% of high-end AI chips from Nvidia, though one prediction was that ratio would fall to about 50-60% over the following five years. China has historically contributed about 20 percent of Nvidia's sales revenue, with 2023 seeing some $1 billion in chips sold to the country.
However, the share of Nvidia's revenue gained from the China / Hong Kong region fell from 21% in October 2023 to 12% in October 2024. This came as Huawei's Ascend 910B 7-nanometer chip rivaled Nvidia's A100 chip, barred from export to China by U.S. control regulations. In August 2024, Huawei was testing its Ascend 910C chip, which Huawei indicated was comparable to Nvidia's H100 4-nanometer AI chip. Mid-April 2025 saw Huawei announce its plan to produce Ascend 920 AI chips as an alternative to Nvidia's H20, which had seen export controls applied earlier in the month.
By mid-July, the Trump administration had reversed course on the Nvidia chip. The resumption of H20 exports to China meant Nvidia would avoid a loss of up to $5.5 billion in sales. At the same time, AMD was cleared to resume sales to China of its MI308 AI chip. While these developments seemed to be a victory for China, and came as part of a trade agreement with the US on rare earth magnets, Lutnick claimed the US still was not selling China “our best stuff.”
This was not the end of the story, in China's eyes. In late July, China's Cyberspace Administration summoned Nvidia representatives over suspicions of “tracking and positioning functions” in the H20 chips. The Chinese government office stated “Recently, Nvidia's computing chips were revealed to have serious security issues.”
In reversing policy shifts, the U.S. seemed to be struggling to formulate a coherent response to China's rapid rise in chip technology. Following its successful global dominance of legacy chips, China looked to be moving forward with its “Delete A” policy, while taking advantage of American components as it refined its own more advanced chip technologies. Should companies like Huawei and other domestic technology corporations succeed in manufacturing cutting edge AI chips, American companies may be left out of vital world sales markets. China's industrial policy could prove to be the key to the nation's semiconductor sector success.
