Inside China’s All-Out Push to Catch Up With American AI Chips
Xi confidant leads fevered effort for domestic alternatives to win AI race; Chinese AI companies warned that anyone who resisted using local chips was a traitor
July 23, 2026 The Wall Street Journal
Last summer, top Chinese chip developer Huawei Technologies held a closed-door briefing for the country’s technology czar to introduce its latest artificial-intelligence chips. The company laid out a plan for rivaling American behemoth Nvidia and said China could become self-sufficient in some critical areas of AI within three years.
It was just what Ding Xuexiang, a confidant of Chinese leader Xi Jinping, wanted to hear. Three years earlier, Washington had squeezed China’s access to cutting-edge AI chips and the tools required to manufacture them—a potentially crippling blow to the country’s ambitions as a tech power. With Xi’s blessing, Vice Premier Ding started a fevered effort to forge domestic alternatives.
Ding dusted off the same all-out approach China had used to produce its first atomic bombs, hydrogen bombs and satellites in the 1960s during a rift with the Soviet Union. He set up a committee that drew from the country’s best companies and labs to form specialized teams and directed them to master the different elements of the chip supply chain.
Huawei’s report on its progress gave Ding leeway to be more aggressive. He delivered a blunt warning to China’s largest AI users, according to people familiar with the message: Anyone who resisted employing domestic chips was a traitor.
In May, Huawei went public with more details of its plans. The company already led the way in reducing China’s dependence on foreign AI chips from 90% in 2021 to less than 60% by 2025, and the newer designs could help lower that number to 25% over the next half-decade, according to data from Morgan Stanley. What’s more, Huawei said it had figured out workarounds for making near state-of-the-art silicon without leading-edge machinery.
“If the U.S. hadn’t forced our country, our company and our industry into a corner, we would never have done something like this,” said Huawei deputy chairman Eric Xu.
In AI software, China is giving the U.S. an unexpected run for its money. Last week, Beijing-based startup Moonshot AI unveiled a new model, Kimi K3, that closely trails the performance of top models from American AI giants OpenAI and Anthropic. Xi delivered a speech positioning China as the global champion of open, accessible AI. Shares in U.S. chip makers fell.
In hardware, however, China still lags far behind the U.S. even with Huawei’s recent progress, and chip-production capacity problems are preventing Chinese companies from deploying AI as quickly as they would like. Moonshot had to pause sign-ups for premium services just two days after launching K3 because it didn’t have enough computing capacity.
The outcome of China’s effort to catch up has ramifications for the global economy. If the U.S. controls the top technology, the rest of the world will depend on Washington’s favor to make progress in fields such as medicine, robotics and autonomous driving—not to mention weaponry and military strategy. AI advances depend on hardware, in particular the chips that train AI models and help them tackle everyday tasks.
A record of overtaking the West in electric vehicles and batteries gives many in China confidence the country can eventually break the U.S. stranglehold on chip technology. But chips are different from other hardware. The technology is extraordinarily complex, and it is moving at a lightning pace.
“It’s trying to catch up with a bullet train,” said Kyle Chan, a fellow at the Brookings Institution. Even at “full China speed,” he said, it’s a daunting challenge.
Ding’s committee bets that by hitching China’s engineering prowess to the engine of state capitalism, it may eventually get there. The push has made progress that few thought possible back in 2022, when Huawei was starved of U.S. technology and Chinese AI companies depended almost entirely on Nvidia.
China’s State Council Information Office didn’t respond to requests for comment.
Executives at Chinese tech companies described a change in mindset over the past year. They had previously dismissed local chips as impossibly far behind. Now they think they can do business with Huawei and other local chip makers. Tech company Meituan recently released an AI model that it said was comparable to Google’s flagship model and was trained solely with Chinese chips.
In memory chips, another key AI need, Chinese companies that barely had sales a few years ago can now catch a glimpse of the American and South Korean companies at the front of the pack.
Chinese chip makers aim to boost the production of advanced wafers used to make chips to more than 500,000 a month by 2030, up from roughly 30,000 last year, according to people familiar with the target. Huawei expects to ship around 1.5 million AI chips this year, roughly doubling its 2025 volume, some of them said.
Still, the technology gap remains considerable. Nvidia’s top AI chip boasts around four times the computing power of Huawei’s best offering. China’s AI computing power stood at around 14% of the U.S. in 2025, research house Bernstein estimated. Despite China’s aggressive build-out, Bernstein expected China would remain significantly behind the U.S. through 2030.
Under Ding’s watchful eye, Chinese companies are using hundreds of thousands of Nvidia chips to stay within striking distance of their American rivals for now.
All out or die
In 2022, the Biden administration instituted export curbs that aimed to slow China’s progress in AI for military uses. They blocked China from purchasing advanced AI chips and tightened access to the top-end, ultraviolet lithography machines and other tools needed to produce them.
Communist Party leaders saw the controls as a threat to its capacity to boost economic growth and compete globally with the U.S. They didn’t yet grasp the promise of large language models like ChatGPT, which OpenAI had just released, but they saw dominance of AI in its many forms as critical to China’s security and the party’s legitimacy.
The chip committee headed by Ding, a 63-year-old who trained as a metalworking engineer, kicked off in early 2023. It identified critical fields—such as advanced manufacturing and packaging, high-bandwidth memory and chip-design software—and recruited top teams in each. They also started deciding which companies would get access to the nation’s limited supplies of chips.
China’s earlier efforts to build a self-reliant domestic chip industry had been marred by corruption and hobbled by bureaucratic mandates. Officials were determined to correct those mistakes. The government encouraged chip-related companies to pursue stock-market listings so they could raise more capital and be subjected to market discipline. Even Huawei would face competition from cloud-computing companies developing their own AI chips.
The government scrapped state salary caps for state-owned equipment maker Naura Technology, allowing the company to lure top talent from foreign rivals.
The Ding plan had a rough first year. Dozens of homegrown AI chips were available on the market, but a government-backed research lab told Premier Li Qiang that they were “very unstable,” according to state media. None were good enough to train large-scale models.
Beijing worried about the reliance of the country’s AI developers on Nvidia. Chinese engineers were locked into Nvidia’s software tools, which could barely be used with domestic chips. The American chip giant had tailored products for China that sat right at the performance limit set by U.S. export controls. Chinese firms also bought banned Nvidia chips from middlemen who routed shipments through other countries.
China’s central economic planning agency brought in several academics and chip experts to help run the project, an unusual move for top government bodies which usually keep decision-making in-house. In 2024, China raised about $48 billion in a state fund to invest in its semiconductor industry.
Turning point
The turning point came last year. Huawei readied a new chip, the Ascend 950, that surprised early testers with its performance. Others, including cloud-computing and AI giant Alibaba, also reported promising progress with their own chips.
Around that time, Nvidia Chief Executive Jensen Huang flew to Beijing with approval from President Trump to sell a downgraded chip designed for China called the H20. In Washington, Commerce Secretary Howard Lutnick said on television that letting China buy Nvidia’s “fourth-best product” would keep the country hooked on American technology.
Ding and other senior officials felt insulted, people familiar with the matter said. Authorities privately told companies to halt their Nvidia purchases, while officials publicly raised cybersecurity concerns about the U.S. company’s products.
Nvidia wanted to win back Beijing. In December, Trump said he told Xi he would allow sales of the company’s more powerful H200 chip. But in Beijing, the calculation had shifted. Chinese officials called in tech companies—sometimes weekly—to evaluate how vital the H200 truly was. In January, Beijing told some companies they could buy the H200s only when necessary, while demanding a commitment to use more domestic chips.
The squeezing of Nvidia boosted demand for domestic chips, creating a supply crunch as Chinese chip makers struggled to ramp up production. It also led to a new boom in the gray market.
Over the past six months, prices for smuggled Nvidia chips have more than doubled, according to Chinese distributors, due to tightened U.S. enforcement and shipping delays in the Middle East. Some Chinese companies have managed to amass clusters of Nvidia’s cutting-edge Blackwell chips. Others are leasing computing power from foreign data centers.
Tech workarounds
Huawei’s efforts to work around such technology restrictions date to 2019. The company had designed an Ascend AI chip that, on paper, matched Nvidia’s top chip at the time, but it never got to produce the chip en masse. Weeks later, the first Trump administration blacklisted Huawei’s chip unit, cutting off its access to U.S. chip technology.
That meant the Chinese company could no longer work with Taiwan Semiconductor Manufacturing, the only chip maker capable of mass-producing the Ascend. He Tingbo, chief of Huawei’s chip division, called it her unit’s “darkest hour.”
Since then, Huawei has increasingly turned to alternative designs, advanced packaging and networking technology to squeeze computing power from older machinery—a practice industry insiders dub “diaohua,” or fine-carving.
Leading-edge chips are made using extreme-ultraviolet, or EUV, lithography systems. The machines use focused beams of ultraviolet light to etch on wafers of silicon billions of transistors roughly the width of a strand of DNA. Tinier processes allow for smaller transistors, meaning more circuitry can be squeezed onto a single sliver of chip to boost computing power. The cutting-edge 3-nanometer process can fit 250 million to 300 million transistors in one square millimeter of silicon.
Because of U.S. export curbs, China has access only to previous-generation deep-ultraviolet, or DUV, machines, which can’t etch with as much precision. Huawei has developed a multistep method for producing 7-nanometer chips using DUV, but it complicates circuit layouts, introduces more opportunities for error and slows output.
The Wall Street Journal reviewed some 200 patents published over the past two years by Huawei and its partners. Those patents illustrate their efforts to reduce errors and boost yield from older machinery through production workarounds and software, sometimes using AI.
One patent describes a wafer-holding plate that can detect and adjust surface temperatures to correct flaws caused by old machinery. It isn’t clear how many of those patented processes are in use on the factory floor.
Huawei revamped the design for its latest Ascend 950 chip, making it more cost-effective. Chinese AI companies including DeepSeek helped Huawei make the chip easier to use. They said it has improved significantly over previous generations and can now effectively substitute for Nvidia chips on certain tasks—particularly inference, where trained models respond to user queries. That is persuading more Chinese companies to use it for AI applications such as the AI assistants.
Huawei also seeks to parlay decades of networking expertise from its telecommunications business into an edge that some say Nvidia has yet to match. The company is attempting to bundle hundreds of thousands of AI chips into massive clusters that function as a single computer to mimic cutting-edge silicon.
“They’ve managed to make up for the lack of access to great chips, critical inputs, by improving much more on their ecosystem efficiency,” Hong Kong University of Science and Technology economist Keyu Jin said at a recent Wall Street Journal event. “This is what China’s truly good at.”
An immediate bottleneck is output. Huawei’s planned production of its new chip—around 750,000 this year—will fall well short of demand. Huawei’s bundling strategy also means it needs several times as many chips to match the computing power of Nvidia’s systems. The company said it was working to increase supplies.
Beijing has encouraged chip companies, including Huawei, to license their workarounds and technology so that they could learn from each other, people familiar with the matter said.
Huawei unveiled its latest “fine-carving” alternative in May. Instead of trying to make transistors smaller, the company laid out a method for stacking circuits to cram more computing power into a single chip footprint. Huawei said that method would allow it to develop chips by 2031 that could outperform any semiconductor currently in production.
“This isn’t about Huawei overtaking the U.S. on a hairpin turn,” said Qingyuan Lin, a semiconductor analyst at Bernstein. “It’s more about finding a way to prevent the gap from widening quickly.”
To truly compete without reliance on foreign technology, China will have to build its own EUV machines, analysts say.
China is likely to figure out how to match the technology eventually, though it will take at least a decade and maybe as many as 50 years, according to Ray Wang, an analyst with research firm SemiAnalysis.
The difficulty lies in the technology’s combination of extreme complexity with extreme precision, said Chris Miller, a historian of technology at Tufts University and author of “Chip War.” In other words, the approach China used to build its first nuclear bomb may not be enough this time around.
“It’s a lot harder than nuclear weapons,” Miller said. “It’s orders and orders of magnitude more complex than almost every other type of manufacturing.”
Copyright ©2026 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8