America's China Tech Decoupling: Semiconductors, AI, and the New Cold War

For three decades, the technology industries of the United States and China operated in a state of deep interdependence. American companies designed chips, Chinese factories manufactured them. American platforms built global user bases, Chinese internet companies built parallel systems. American researchers published papers, Chinese universities trained students on them. That era is over. The decoupling now underway is not a policy debate — it is an accomplished fact being executed at scale, and its consequences will reshape the global economy more profoundly than any trade dispute in history.

The Semiconductor Chokepoint

The strategic logic of the US approach to China's technology sector rests on a single observation: advanced semiconductors are the indispensable input to modern military capability, economic productivity, and AI development. Every sophisticated weapons system, every frontier AI model, every advanced communications network runs on chips that China cannot currently manufacture for itself. This dependence creates leverage, and Washington has moved aggressively to exploit it.

The Commerce Department's export control regime, initially implemented in October 2022 and substantially expanded in subsequent rounds, now restricts China's access to chips manufactured with American equipment, American software, or American intellectual property — a reach that covers virtually the entire global semiconductor supply chain due to the dominance of American companies like Applied Materials, Lam Research, KLA Corporation, and Cadence Design Systems in chip manufacturing tools and design software. The rules have been extended to cover third-country manufacturers, meaning TSMC in Taiwan and Samsung in South Korea cannot sell advanced chips to Chinese customers using American-origin technology.

The practical effect has been significant but incomplete. China's Huawei, long the primary target of US chip restrictions, managed to release a 5G smartphone in 2023 powered by a 7-nanometer chip fabricated domestically by SMIC — a technical achievement that caught US officials off guard and demonstrated the limits of export controls as a complete solution. Subsequent tightening of the rules has made further advances more difficult, but the episode established that determined adversaries with sufficient resources can achieve significant progress despite restrictions.

TSMC and the Taiwan Question

No element of the tech decoupling is more geopolitically fraught than the status of Taiwan Semiconductor Manufacturing Company. TSMC manufactures approximately 90% of the world's most advanced chips, and its fabs sit 100 miles from mainland China. The strategic vulnerability this represents has prompted an unprecedented effort to diversify production away from the island.

TSMC's Arizona facility, supported by $6.6 billion in grants and $5 billion in loans under the CHIPS and Science Act, is now in production — though at costs significantly higher than Taiwan operations and with yields that have required extensive engineering effort to optimize. Intel's ambitious foundry strategy, backed by $8.5 billion in CHIPS Act funding, has faced setbacks that have shaken confidence in America's ability to quickly rebuild domestic manufacturing capacity. Samsung's Texas facility adds a third node, but the aggregate capacity of US-based advanced chip manufacturing remains a fraction of global demand.

The strategic calculus has also prompted TSMC to announce a Japan facility, ASML — the Dutch company that holds a monopoly on extreme ultraviolet lithography equipment essential for advanced chips — to restrict China sales under Dutch government pressure, and a wave of similar measures across the allied technology supply chain. The result is a nascent "trusted supply chain" architecture that mirrors Cold War-era COCOM export control regimes in its ambition, if not yet its completeness.

China's Response: Self-Sufficiency or Bust

Beijing has responded to the decoupling pressure with the most ambitious industrial policy effort in modern history. The Big Fund — China's national semiconductor investment vehicle — has deployed over $47 billion across chipmakers, equipment manufacturers, and materials suppliers. The "Made in China 2025" successor programs target domestic supply chain sufficiency in semiconductors, AI, quantum computing, aerospace, and biotechnology.

Progress has been uneven. In some areas — electric vehicles, solar panels, battery technology — China has achieved genuine global leadership. BYD surpassed Tesla in global EV sales in 2023 and has not looked back. CATL dominates global lithium-ion battery manufacturing. These achievements reflect genuine industrial policy effectiveness and should not be dismissed as mere subsidies without technological substance.

In semiconductors and advanced AI, the picture is more complicated. China has world-class AI researchers and companies — Baidu, Alibaba, and Huawei have each built large language models that perform credibly on Chinese-language tasks. But the export controls on training compute have materially constrained Chinese AI development relative to American counterparts, a gap that Chinese researchers themselves acknowledge in private conversations reported by academics at Stanford's Hoover Institution. The question is not whether China can develop AI, but whether it can develop frontier AI at the pace of its American competitors without access to the necessary hardware.

The Investment Decoupling

Capital flows have tracked the technology restrictions. US venture capital investment in Chinese tech companies fell from a peak of $35 billion in 2021 to under $5 billion in 2025, according to data from PitchBook. The Biden administration's outbound investment executive order, formalized into regulation in 2024, restricts American investment in Chinese companies working on advanced semiconductors, quantum information technology, and AI systems with potential military applications.

Chinese investment in American technology companies has been under scrutiny from the Committee on Foreign Investment in the United States since the late 2010s, but CFIUS reviews have become substantially more aggressive. Several proposed acquisitions of American semiconductor design companies by Chinese-linked entities have been blocked, and existing Chinese investments in companies ranging from TikTok parent ByteDance to security camera manufacturers have faced forced divestiture orders.

The financial decoupling creates particular complications for American pension funds, university endowments, and sovereign wealth funds that have historically invested in Chinese technology companies through index funds and direct holdings. The MSCI China Index, which many passive investors track, contains substantial allocations to Chinese tech companies that would fall within the scope of contemplated investment restrictions — a fact that has generated considerable anxiety among institutional investors navigating competing regulatory and fiduciary obligations.

The Road Ahead

The decoupling is irreversible in the near term. The political consensus in Washington for restricting Chinese access to advanced technology spans both parties in a way that few policy questions do. The national security rationale — that advanced technology capabilities translate directly into military power that could be used against American interests — is not contested in any serious foreign policy forum.

What is contested is the efficacy and cost of the decoupling strategy. Critics including former Treasury Secretary Larry Summers have argued that export controls leak, that determined adversaries find workarounds, and that the economic costs to American companies — which are also losing access to the Chinese market in response to US restrictions — are being systematically underestimated. Supporters argue that even imperfect controls buy time, and that time has strategic value when the race for AI supremacy is measured in months.

The deeper question is whether the world's two largest economies can sustain separate, incompatible technology ecosystems indefinitely — or whether the economic gravity of interdependence will eventually reassert itself. History suggests that economic forces ultimately prevail over political ones, but history has never seen a technology competition of this scope or strategic consequence.

For additional context, see our analysis of domestic US AI regulation and the AGI race between leading American labs that drives much of the geopolitical competition.