By Pranav Chopra
The world’s biggest trade war is not being fought over oil, steel or rare earth minerals. It is being fought over tiny semiconductor chips that power everything from smartphones and electric vehicles to cutting-edge artificial intelligence (AI) systems and modern weapons. With AI becoming increasingly indispensable for economic growth, military power and technology leadership, controlling the advanced semiconductor industry has become one of the key strategic battles of the twenty-first century.
The United States has spent the last four years trying to slow China’s technological rise by limiting its access to the world’s most advanced semiconductors and the specialized equipment required to make them. The strategy was meant to keep America ahead of the curve in AI and other critical technologies. However, the restrictions have not held China back, but indeed spurred Beijing’s drive for semiconductor self-sufficiency, all while costing American companies: most notably Nvidia has lost billions in sales. The result is a technology war that is reshaping global supply chains and creating two increasingly detached semiconductor ecosystems.
The US Builds a Wall
Major change occurred in October 2022, when the Biden administration announced new broad export restrictions on advanced integrated circuits, computers containing them and the specialized equipment used to build them. The rules were aimed at preventing China from obtaining the chips needed to train and run the most advanced artificial intelligence (AI) systems and were tightened in October 2023 and again in December 2024. The goal of the controls was clear: stop China from buying today’s best chips and prevent it from making tomorrow’s. Many people believed this would leave China several years behind the United States and its allies.
The logic behind this strategy depended heavily on the technological gap between China and the United States and its allies, particularly in the ability to manufacture increasingly advanced chips. That confidence rested on a simple technical yardstick: the size of the chips, measured in nanometres (nm). The smaller the number, the more powerful and efficient the chip. Making these chips requires expensive factories called fabs and extremely advanced machines that only a handful of companies in the world can produce.
China responded faster than expected. In 2023, Huawei stunned Washington by unveiling a smartphone built on a 7-nanometre chip made domestically by SMIC, China’s biggest chip manufacturer – something many experts had thought the restrictions would make impossible. Washington reacted by tightening the rules again: Nvidia kept designing weaker versions of its chips for the Chinese market, but in April 2025 even those were blocked. The company said the license requirement would apply indefinitely, and it went on to book a $4.5 billion charge tied to unsellable H20 inventory for its Chinese customers.
China Speeds Up
This is where the policy began to generate unintended consequences. Although China had pursued semiconductor self-sufficiency for years, the export restrictions significantly increased the urgency of that objective. In May 2024 Beijing launched the third phase of its state-backed “Big Fund”, registered with 344 billion yuan (roughly $47.5 billion) – the largest of the fund’s three rounds since 2014.
More importantly, Chinese companies suddenly had a guaranteed market. Beijing has pushed state-linked buyers toward domestic suppliers, and by November 2025 it was directing newly built, state-funded data centres to use only Chinese-made AI chips. That gave homegrown chipmakers something every new industry needs: guaranteed customers. The results appeared quickly. Cambricon, often nicknamed “China’s Nvidia” reported 2025 revenue up 453% year-on-year and its first-ever annual profit turning around a loss from the prior year. Of the roughly four million AI accelerator cards shipped in China in 2025, domestic vendors supplied about 1.65 million with Huawei alone delivering more than 800,000. Nvidia, which once controlled close to 95% of China’s AI chip market saw its share fall to roughly 55%, according to research firm IDC as reported by Reuters. Taken together, these figures suggest that the restrictions did more than constrain China’s access to foreign technology: they helped create the scale, demand and commercial incentives needed for domestic competitors to develop. Even the Center for Strategic and International Studies, an advocate of tougher controls, has acknowledged the policy accelerated China’s use of homegrown technology.
China Still Faces Big Challenges
Despite this progress, China is still behind the world’s leading chipmakers. SMIC’s most advanced logic chips are stuck at a 7-nanometre-class process, and by late 2024 its newest smartphone chip for Huawei had only crept marginally beyond that node. Meanwhile Taiwan’s TSMC started volume production of 2 nanometre chips, several generations ahead in the fourth quarter of 2025.
China also lacks the world’s best lithography machines forcing SMIC to rely on slower, costlier production methods. That contributes to low yields, the share of usable chips produced from each silicon wafer. In early 2024, there were reports that the yield on Huawei’s Ascend 910B processors was only around 20% and by early 2025, the Financial Times reported that figure had almost doubled to about 40% for the newer 910C. Still, most global chipmakers target yields well above 70% to reliably make a profit.
Another major weak spot is memory. Modern AI chips depend on high-bandwidth memory (HBM) and research firm SemiAnalysis has suggested this remains one of Huawei’s biggest constraints on scaling production. A teardown by TechInsights also found that some of Huawei’s chips still contain processor dies originally made by TSMC years before export controls tightened, raising questions about how long stockpiled components can sustain output.
In June 2025, US Commerce Under Secretary Jeffrey Kessler told Congress that Huawei’s Ascend chip production would likely stay at or below 200,000 units for the year, a fraction of the millions of AI processors Nvidia ships annually – though he cautioned lawmakers against complacency, saying China was “catching up quickly”.
Policy Begins to Change
Over time, even Washington started questioning whether a total blockade was the right approach. Some officials worried that shutting Nvidia out entirely was simply handing Huawei a captive market, and parts of the policy were relaxed as a result.
In August 2025, Nvidia and AMD agreed to pay 15% of their Chinese revenue to the US government in exchange for export licenses for the H20 and MI308 chips. In December 2025, Washington went further, approving exports of the more powerful H200 chip subject to a steeper 25% levy. What began as an attempt to deny China access to advanced chips was becoming something more complicated: a managed trade-off in which American firms could sell but only under increasingly costly conditions. The irony was that by the time Washington began reopening the door, the pressure of the earlier restrictions had already pushed Chinese firms and buyers towards alternatives.
By early 2026, Chinese customs authorities were blocking H200 shipments outright even after US approval, with Beijing steering investment toward Huawei and other domestic suppliers instead. US Commerce Secretary Howard Lutnick confirmed to lawmakers in April 2026 that not a single H200 had been sold in China as a result. Nvidia CEO Jensen Huang has described the company’s China market share as having collapsed from 95% to essentially zero and has acknowledged Nvidia has largely ceded the Chinese market to Huawei.
Perhaps the clearest symbolic moment came in April 2026, when DeepSeek released its V4 model built to run on Huawei’s Ascend chips rather than Nvidia hardware – a first for one of China’s frontier-class AI labs and a sign of how quickly the country’s AI industry is adapting to domestic silicon.
Conclusion
The semiconductor trade war has demonstrated that economic sanctions can reshape industries in unexpected ways. The United States has succeeded in making China’s path to the most advanced chips more difficult with companies such as SMIC and Huawei still facing significant technological hurdles. Yet the same restrictions have also accelerated Beijing’s efforts to build a self-reliant semiconductor industry, encouraging massive state investment and reducing China’s dependence on foreign suppliers. Rather than producing a clear winner, the export controls have transformed the competition itself, shifting it from a race for technological leadership into a long-term contest over industrial resilience and innovation.
The result is a more divided technology industry. Rather than depending on the same global supply chains, both countries are increasingly constructing separate systems. The race is no longer only about who can make the fastest chip. It is also about who can keep innovating, produce at scale, and sustain a strong domestic industry over the long run. Four years after the first export controls were introduced, one thing is clear: the Great Wall of Silicon is still being built and neither side has won yet.
The views expressed in this article are the author’s own and may not reflect the opinions of The St Andrews Economist.
Image Credit: Unsplash