Four years after the United States tightened export controls on advanced semiconductors used in AI, data centers and defense, Beijing has accelerated a long-standing push for chip self-reliance. What began under the Made in China 2025 program has been reinforced by hundreds of billions of dollars in state support — subsidies, tax breaks and incentives aimed at building domestic rivals to US and Taiwanese leaders like NVIDIA and TSMC.
Those policies have helped scale up capacity. The mainland foundry SMIC reported record revenues of $9.3 billion last year, and HuaHong has been operating above capacity amid surging demand. Still, most outside experts say China is not yet fully self-sufficient. Observers such as Ryu Yongwook of the National University of Singapore note that China trails the US in research, design and innovation, and lags Taiwan and South Korea in the most advanced production techniques.
China has, however, made important gains on older process nodes. The Rhodium Group estimates Chinese firms now control roughly 30% of the global market for legacy chips — components widely used in cars, industrial equipment and many consumer devices. Large-scale production of these less advanced but essential semiconductors has lowered global prices and put pressure on non-Chinese suppliers in areas such as silicon carbide wafers for high-power applications.
On cutting-edge nodes the picture is more mixed. China has produced 7-nanometer-class processors that power recent Huawei smartphones — performance roughly comparable to TSMC chips from around 2018, but behind the 3nm and 5nm designs that lead on speed, power efficiency and cost. Analysts warn that material shortages, supply-chain limits and US sanctions mean catching up to the latest process nodes could take a decade or more.
Policy signals from Beijing also suggest a shift in emphasis. The Communist Party’s recent Five-Year Plan places less explicit focus on chip dominance and highlights AI within a “model-chip-cloud-application” framework that treats advanced semiconductors as part of a broader computing stack. The government appears to be prioritizing pragmatic, task-oriented AI for industry — workloads that often require less raw compute and that domestic chips can handle more cost-effectively.
That cost advantage matters. Chinese chips and AI systems frequently trade peak performance for much lower price, which has boosted adoption across the Global South where governments and firms increasingly choose more affordable Chinese solutions. Trendforce reported that Chinese AI platforms — including DeepSeek and Alibaba’s Qwen — captured roughly 15% of the global AI model market by late 2025. That expansion represents a long-term competitive challenge for big US cloud and AI vendors even as those firms commit record spending to AI infrastructure.
Practical limits also constrain the US edge. ICIS has warned that US data centers, which rely on high-end chips, could be hampered by strains on the national power grid. By contrast, China’s expanding power sector could provide spare capacity for large-scale data-center deployment; ICIS projects roughly 400 gigawatts of spare capacity in China by 2030. Lower energy costs can help offset relative chip inefficiency for compute-heavy workloads.
ICIS outlines three broad scenarios: the US repairs grid weaknesses and keeps its lead; the US continues to lead advanced chip research while Chinese AI systems spread across the Global South; or geopolitical and trade tensions split the market into two separate AI ecosystems. Whichever path unfolds, established chipmakers face a future where lower-priced Chinese competitors are narrowing gaps in sophistication and reliability.
Edited by: Tim Rooks