The Chip War: Semiconductors & Export Controls
The competition for advanced semiconductors has become the most tangible front of the technology cold war. Control over chip design, manufacturing, and supply chains now sits at the intersection of economic strategy, industrial policy, and national security.
The Chip War, the competition for advanced semiconductors, has become the most tangible front of the technology cold war. Control over chip design, manufacturing, and supply chains now sits at the intersection of economic strategy, industrial policy, and national security.
Why Chips Are the Defining Battleground
Semiconductors underpin virtually every domain of modern technology competition — from artificial intelligence and autonomous systems to telecommunications, defence, and advanced manufacturing. Without access to cutting-edge chips, frontier AI models cannot be trained at scale. Without the fabrication capacity to produce them, nations remain dependent on supply chains they do not control.
This dependency is not theoretical. It is the strategic reality that has driven the United States to impose the most significant technology export controls since the Cold War — and that has driven China to mobilise state resources, private capital, and academic research toward semiconductor self-sufficiency at an unprecedented scale.
The chip war is not a metaphor. It is an active, high-stakes contest over the physical infrastructure of the digital economy.
The Architecture of the Chip War
The global semiconductor supply chain is extraordinarily concentrated. A handful of companies and a single geography occupy chokepoints that no major power can easily replicate or bypass.
Taiwan Semiconductor Manufacturing Company (TSMC) fabricates approximately 90 percent of the world's most advanced chips. ASML, a Dutch company, is the sole manufacturer of extreme ultraviolet (EUV) lithography machines — the equipment required to produce processors at the leading edge. Nvidia designs the GPUs that dominate AI training and inference. Each of these companies represents a single point of strategic vulnerability — and a single point of leverage.
The United States has sought to exploit this concentration through export controls. Beginning in October 2022 and expanded in subsequent rounds, the Bureau of Industry and Security (BIS) has restricted China's access to advanced AI chips, chipmaking equipment, and the technical expertise required to produce them domestically. The objective is to maintain a generational lead — to keep China at least two technology nodes behind the frontier.
Read our full analysis: The Silicon Curtain — How Export Controls Reshaped the Chip Industry →
The Evolving Export Control Landscape
Export control policy is not static. It has undergone significant shifts as both the technology and the geopolitical context have evolved.
In January 2026, the BIS revised its licensing framework, moving from blanket denials to case-by-case assessments for chips such as Nvidia's H200 and AMD's MI325X. The policy allows controlled exports to China subject to a 25 percent security tariff, reflecting a pragmatic calculation: keeping global AI development anchored to American technology while generating revenue for domestic chip subsidies.
The shift has not been without controversy. In Congress, the AI Overwatch Act — which would give lawmakers a 30-day veto over any semiconductor export licence to China — continues to advance. The tension between those who favour commercial engagement and those who advocate maximum restriction remains unresolved.
Meanwhile, Nvidia has navigated the regulatory landscape by designing successive chip variants — the A800, the H800, the H20 — each calibrated to fall just below the performance thresholds set by export controls. This cat-and-mouse dynamic between regulation and product design shows no sign of ending.
China's Response: The Drive Toward Self-Sufficiency
China has not responded to Western restrictions with passivity. Under what Beijing describes as a "New National System," the state has mobilised an integrated effort spanning government, private industry, and academia to build a domestic semiconductor capability.
Huawei has emerged as the most visible symbol of this effort. Its Ascend 910 series represents China's most advanced domestically produced AI processor. While the Ascend chips still lag behind Nvidia's Blackwell generation in raw performance, Chinese engineers have compensated through cluster-scale computing — linking tens of thousands of processors into networked systems that approach the aggregate capability of Western supercomputers.
Beyond Huawei, a new generation of Chinese chip companies — Moore Threads, Biren Technology, MetaX, and Enflame — has emerged to serve the domestic market. These firms are expected to capture a significant share of China's AI chip demand by late 2026, reducing reliance on restricted Western imports.
The deeper strategic question is not whether China can match individual Western chips, but whether it can build a self-sustaining ecosystem — from design tools and fabrication to packaging and deployment — that functions independently of the chokepoints controlled by the United States and its allies.
The CHIPS Act and the Reshoring Bet
The United States has not relied solely on restricting China. Through the CHIPS and Science Act, Washington has committed over $52 billion to rebuilding domestic semiconductor manufacturing capacity.
TSMC's Arizona fabrication facility is ramping up. Intel is investing in new fabs in Ohio and Oregon. Samsung is building in Texas. The objective is to reduce America's dependence on Taiwan — a dependency that carries acute risk given the island's geopolitical vulnerability.
The timeline, however, is measured in years, not months. Matching the output, yield rates, and cost efficiency of Taiwan's existing facilities remains a generational challenge. Industry observers note that meaningful domestic production capacity extends well into the next decade.
The CHIPS Act represents a bet that industrial policy can reshape semiconductor geography. Whether that bet pays off — or whether the commercial logic of Asia-based manufacturing proves too powerful to override — will be one of the defining economic questions of the coming decade.
The Jevons Paradox and AI Demand
One of the more counterintuitive dynamics in the chip war is the relationship between efficiency and demand. DeepSeek's success in training competitive AI models at a fraction of the cost of Western counterparts initially appeared to reduce the need for massive GPU procurement. Nvidia's stock fell sharply on the news.
But the Jevons Paradox — where efficiency gains increase rather than decrease total demand — has so far prevailed. As AI models become cheaper to run, they are deployed more widely, driving aggregate demand for compute upward. Goldman Sachs projects that AI-related infrastructure investment may exceed $500 billion in 2026. Nvidia's valuation has recovered and surpassed its pre-DeepSeek peak.
The implication is that the chip war will intensify, not subside. Even as architectural innovations like DeepSeek's Engram reduce per-model hardware requirements, the expansion of AI into new domains — agents, robotics, autonomous systems — will sustain and likely accelerate demand for advanced semiconductors.
Read more: The AI Arms Race — From GPT to DeepSeek →
What We Cover
Tech Cold War tracks the semiconductor competition across its full scope: export control policy and its consequences, the domestic chip programmes of the US and China, the strategic role of TSMC and ASML, the emergence of Chinese chip alternatives, the reshoring of manufacturing capacity, and the intersection of hardware supply chains with AI development.
This page is updated regularly as new analysis is published. Subscribe to receive our coverage directly.
Tech Cold War publishes independent analysis of the technology competition reshaping global power. Subscribe to receive weekly briefings and deep-dive articles.