Illustration: Chen Xia/GT
The global semiconductor contest has long evolved from mere business competition into a high-stakes geopolitical race.
Recently, Nvidia reportedly has told suppliers to suspend production of its H20 chip. At the same time, it is also reportedly working on a new chip for the Chinese market dubbed the B30A, which will be more advanced than the H20. Nvidia’s setbacks with its H20 chip and its proposed China-specific B30 chip represent merely the tip of the iceberg in this great power rivalry.
On the surface, this appears to be another round of US high-tech containment against China; at a deeper level, it reflects a significant shift in China-US technological competition - from "complete embargo" toward "controlled access" with conditions attached.
In this protracted struggle, the decisive factor lies not in Washington’s policy oscillations but in the pace of China’s technological breakthroughs and indigenous innovation capabilities.
American chip policy toward China exhibits a conspicuous "dual character" - simultaneously striving to maintain technological leadership while reluctant to forfeit the world’s largest market. This contradictory mindset has led to frequent policy reversals.
In April, the US government halted Nvidia’s H20 sales to China; just a few months later, it reversed course by issuing export licenses — with the condition of collecting a 15 percent revenue share.
This shift doesn’t reflect sudden American benevolence but rather economic reality forcing its hand — Nvidia alone suffered losses estimated at $12.5 billion in the Chinese market during the first half of 2024.
Nvidia’s predicament isn’t coincidental but inevitable as the China-US semiconductor rivalry reaches a critical juncture.
Nvidia’s challenges in China neither represent a simple corporate stumble nor the aftershocks of a policy misjudgment — they reflect the visible collision of two forces: the US’ attempt to establish regulatory "ceilings" versus China’s accelerating upward momentum.
The increasing difficulty for Nvidia to access the Chinese market actually signals the growth of domestic alternatives, a forming ecosystem, and the convergence of computing power with applications. What represents market resistance for Nvidia serves as a mirror reflecting progress for China’s semiconductor industry.
This isn’t a technological choice, but rather a geopolitical engineering decision. More critically, this "threshold-based openness" reveals something significant: if catching up posed no pressure, such precisely calibrated thresholds wouldn’t be necessary. The moment thresholds appeared, the pursuit was already underway.
The US’ adoption of the "special edition plus fee" model aims to strike a balance between technological containment and market interests. However, as regulations tighten and entry permits become more expensive, the calculus becomes increasingly complex and requires constant adjustment, and in turn US semiconductor regulations will become trapped in perpetual revision.
Ultimately, the decisive outcome won’t depend on how much the US can profit, but rather how China continuously advances indigenous, multi-layered, multi-directional and multi-nodal breakthroughs.
Markets never erect barriers against inconsequential competitors. Nvidia’s difficulties entering the Chinese market precisely confirm the rise of China’s semiconductor industry.
Compared with giants like Nvidia, domestic AI chips still show apparent gaps in computational density, ecosystem completeness and software optimization. China still faces "chokepoint" vulnerabilities, particularly in advanced processes, EDA tools and high-end materials.
However, analyzing the development trajectory over the past two years reveals that China’s semiconductor industry is forging a new path. First, the Chinese government is driving consolidation in the semiconductor industry, concentrating scattered resources into a few globally competitive enterprises.
Second, domestic AI chips have begun achieving breakthroughs in specific segments. AI companies like DeepSeek are designing optimization standards for next-generation domestic chips, attempting to substitute imported chips in model training and inference domains. Meanwhile, AI technology applications across broader industrial sectors have achieved global leadership positions.
This "focused breakthrough" strategy is showing results, most prominently in achieving self-sufficiency in key areas rather than attempting comprehensive coverage.
The Chinese semiconductor industry’s breakthrough won’t happen overnight. Building a complete industrial chain from design through manufacturing to packaging and testing requires time. But confidence doesn’t come from slogans but from verifiable timelines.
Past trajectories demonstrate that continuous iteration, even in small steps, eventually produces compound returns. External single-source solutions are losing their irreplaceability; and emphasizing limitations reveals anxiety about the pace of catch-up.
While challenges undoubtedly exist, the trend is clear: Self-sufficiency isn’t a passing phase but a path that time will validate. What determines everything isn’t how loud the slogans are, but how quickly tangible technological breakthroughs occur. As long as speed persists, ceilings remain merely temporary barriers.
原文地址：http://en.people.cn/n3/2025/0825/c90000-20356783.html