Silicon Supremacy: The US-China Semiconductor War and the Battle for the Future of Artificial Intelligence

Executive Summary

The contest between the United States and China for dominance of the global semiconductor industry has evolved far beyond a commercial dispute or even a technology rivalry. It has become the defining strategic confrontation of the twenty-first century - the terrain on which military advantage, economic supremacy, and the governance of artificial intelligence will be decided for decades to come. Whoever controls the most advanced chips controls the training of AI models, the autonomy of weapons systems, the processing power of intelligence networks, and the computational backbone of the modern economy. There is no domain of national power - from nuclear command-and-control to financial surveillance, from drone warfare to social management - that does not depend, at its core, on semiconductors.

As of mid-2026, the war is being fought across four simultaneous fronts: export controls and counter-controls, domestic chip industrialization on both sides, diplomatic coalition-building among allied semiconductor nations, and the race for artificial intelligence model capability. The United States has deployed the most sweeping technology denial regime since the Cold War. China has responded with the most ambitious state-directed industrial mobilization in semiconductor history. Neither side is winning decisively. Both sides are paying costs that neither fully anticipated. And the Trump administration's oscillating export control policy - tightening, then loosening, then monetizing access to advanced chips at a 15 percent revenue premium - has introduced a dimension of strategic confusion that is simultaneously enriching Washington's treasury and potentially undermining the multilateral coalition upon which the long-term effectiveness of chip controls ultimately depends.

This report provides a doctrine-level assessment of the semiconductor war's current state, its structural dynamics, the capabilities China has developed under sanctions pressure, the coherence of the American and allied strategy, and the realistic trajectories of technological decoupling over the next six to twenty-four months.

Strategic Background

Semiconductors are unique among strategic technologies. They are both a product and an enabler - a physical artifact that unlocks the functionality of virtually every other technology that matters to modern states. Advanced logic chips at the seven-nanometer threshold and below represent capabilities not merely of commercial importance but of direct military application: autonomous weapons guidance, signals intelligence processing, electronic warfare, nuclear weapon simulation, and the training of large-scale AI systems that will reshape every domain of conflict. When National Security Advisor Jake Sullivan described export controls on advanced chips as a new strategic asset in America's toolkit - and announced that Washington no longer sought to maintain merely a generational lead over China but an absolute technological ceiling - he was articulating a doctrine that broke with forty years of technology export policy premised on engagement and managed competition.

The strategic logic is straightforward. The semiconductor supply chain is simultaneously globalized and concentrated in ways that create natural chokepoints of extraordinary leverage. Chip design is dominated by US companies operating on US-origin intellectual property. The most advanced lithography machines - without which cutting-edge chips cannot be manufactured - are produced exclusively by ASML, a Dutch firm whose extreme ultraviolet systems are subject to export controls maintained through US diplomatic pressure on the Netherlands. Semiconductor manufacturing at the leading edge is concentrated primarily at Taiwan Semiconductor Manufacturing Company in Taiwan, a geopolitical flashpoint in its own right. Key materials, specialty chemicals, and manufacturing equipment come from a handful of firms in Japan, South Korea, and the United States itself. Washington concluded that by coordinating controls at these chokepoints, it could deny China the technological substrate for advanced AI and military modernization - without fighting a conventional war.

China's assessment of its own vulnerability has been equally clear-eyed. Beijing's Made in China 2025 program, launched in 2015, explicitly targeted semiconductor self-sufficiency as a national security imperative. The program invested heavily in domestic chip design, foundry capacity, and equipment development - but fell significantly short of its localization targets, with domestically produced chips accounting for approximately 30 percent of domestic consumption in 2025. The advent of comprehensive allied export controls, rather than deterring China's ambitions, galvanized a far more aggressive and better-funded national mobilization. China's government-backed Big Fund III - a 47.5-billion-dollar investment vehicle launched in May 2024 - represents the most concentrated state capital injection into semiconductor development in history. The war, in this sense, has not slowed China's chip ambitions. It has turbo-charged them.

Historical Context

The US-China technology contest did not begin with chips. It began with Huawei. The Trump administration's first-term decision to place Huawei on the Commerce Department's Entity List in May 2019 - cutting off the world's largest telecommunications equipment manufacturer from access to American technology, including the chip designs that powered its smartphones and 5G infrastructure - was the opening salvo in what would become a comprehensive technology denial campaign. The action was both symbolically and practically significant: it demonstrated that the US was prepared to weaponize the global technology supply chain against specific Chinese entities, regardless of the commercial cost to American firms or the diplomatic friction with allies that depended on Huawei's equipment.

The October 2022 export controls imposed by the Biden administration represented a qualitative escalation. The controls went beyond targeting specific entities and imposed technology-wide restrictions on the transfer of advanced computing chips, semiconductor manufacturing equipment, and design software to China. They deployed the Foreign Direct Product Rule - an instrument that extends US jurisdiction to foreign-made products incorporating American technology - to reach chipmakers in South Korea, Taiwan, and elsewhere, compelling them to stop supplying China's most advanced fabrication facilities. The controls established an explicit threshold: logic chips with architectures below 16 nanometers, memory chips below 18 nanometers - the frontier of capability that China desperately needed to advance its AI and military programs.

The October 2022 measures were tightened in October 2023 and again in December 2024, closing loopholes that Chinese firms had exploited through subsidiary restructuring, shell company procurement, and the engineering of chips calibrated to fall just below control thresholds. NVIDIA's H800, then its H20, were each engineered specifically to comply with successive rounds of controls - and each was eventually restricted as the controls tightened around them. Huawei itself was found to have evaded controls at scale: forensic teardowns of its Ascend 910B and 910C chips revealed that the chips contained dies manufactured at TSMC's seven-nanometer node, procured through a Cayman Islands-registered shell company called Sophgo. Approximately 2.9 million TSMC seven-nanometer dies were acquired through this evasion scheme. TSMC was fined one billion dollars for the violation - a penalty that underscored both the scale of the evasion and the systemic difficulty of enforcing controls across a globally integrated supply chain.

The historical analogy that US strategists invoke is Cocom - the Coordinating Committee for Multilateral Export Controls, the Cold War-era Western alliance that restricted technology transfers to the Soviet Union. The analogy is imperfect: China is far more deeply integrated into the global economy than the Soviet Union ever was, and the semiconductor supply chain runs through allies - South Korea, Japan, the Netherlands, Taiwan - that have substantial independent commercial interests in the Chinese market. But the structural logic is similar: use technology chokepoints to deny a strategic competitor the industrial capacity for military modernization, while building the Western industrial base to widen the technological gap.

Current Situation Assessment

The semiconductor war as of mid-2026 is characterized by a fundamental paradox in American strategy. The Biden administration constructed the most comprehensive technology denial architecture in modern history - and the Trump administration has been systematically loosening it in exchange for diplomatic concessions, commercial revenue, and trade negotiating leverage. The contradiction is not subtle. In April 2025, the Trump administration banned even the NVIDIA H20 - a chip specifically engineered to fall below export control thresholds - from sale in China. By July 2025, it reversed course, allowing H20 sales again. By August 2025, it approved NVIDIA's H200 and AMD's MI308 for sale to China on a case-by-case basis. By the time of Trump's May 2026 state visit to Beijing alongside NVIDIA CEO Jensen Huang, the United States had cleared approximately ten major Chinese tech companies - including Alibaba, Tencent, and ByteDance - to purchase up to 75,000 H200 units each, subject to a 15 percent revenue fee payable to the US Treasury. Chinese firms had placed orders for over two million H200 units, worth an estimated 14 billion dollars.

The strategic confusion generated by this oscillation is substantial. Allies - particularly the Netherlands, Japan, and South Korea - have aligned their own export control regimes with US policy at significant commercial cost. ASML has warned investors that its 2026 China revenue will decline materially as a result of control compliance. Japanese semiconductor equipment firms have similarly absorbed revenue losses. When Washington then loosens its own controls while receiving revenue from the relaxation, allied governments face questions from their domestic industries about why they should continue bearing the economic burden of a policy that Washington itself is not consistently enforcing. The institutional architecture of multilateral chip controls - painstakingly negotiated through what analysts call the Chip4Alliance framework encompassing South Korea, Taiwan, Japan, and the Netherlands - is under strain precisely when the long-term effectiveness of the strategy depends most on its cohesion.

The other defining feature of the current situation is the emergence of a genuinely bifurcated AI chip ecosystem within China. TrendForce projects that in 2026, the domestic share of China's AI chip market will reach 50 percent - an extraordinary shift from near-total foreign dependence five years earlier. This is not primarily because Chinese chips have achieved parity with American ones. It is because state-mandated adoption models - in which Chinese cloud companies and AI laboratories are incentivized or required to deploy domestic alternatives regardless of performance gaps - are creating a protected domestic market of sufficient scale to sustain continued development. Alibaba has reportedly delivered over 100,000 units of its Zhenwu 810E AI accelerator, claimed comparable to NVIDIA's H20. At least nine Chinese AI chip companies have exceeded 10,000 shipments or total orders. The domestic AI chip ecosystem, while not yet competitive with American frontier hardware on raw performance metrics, is becoming commercially self-sustaining in ways that purely performance-based assessments fail to capture.

Power Center Analysis

The United States: The Distracted Hegemon

Washington's structural advantages in the semiconductor war remain formidable. American firms hold dominant positions in chip design - NVIDIA, AMD, Qualcomm, Intel, Apple's silicon division - and in the electronic design automation software without which no modern chip can be developed. The intellectual property embedded in every advanced chip produced anywhere in the world traces, at some point, to American-origin technology, which is precisely why the Foreign Direct Product Rule provides such extraordinary extraterritorial reach. The CHIPS and Science Act, signed in 2022, committed 52 billion dollars to domestic semiconductor manufacturing incentives and research, with TSMC, Samsung, and Intel constructing or expanding fabrication facilities in Arizona and Texas.

The structural weakness is strategic coherence. The Trump administration has used chip export controls as a bargaining chip in tariff negotiations, a revenue instrument through the 15-percent fee mechanism, and a diplomatic concession in US-China summit diplomacy - functions that are each individually defensible in narrow terms but collectively corrosive to the signal that an effective long-term denial strategy requires. When controls are perceived as transactional rather than principled, adversaries learn to buy their way around them. The AI Overwatch Act, passed by Congress in January 2026, introduced a mechanism allowing the legislature to revoke Commerce Department-issued export licenses at any time - a provision that creates permanent uncertainty for both American chip companies and their Chinese customers, and undermines the stable regulatory environment that the CHIPS Act investments depend on.

China: The Sanctioned Innovator

China's response to the semiconductor war is best characterized as sanctioned innovation - the phenomenon by which denial of access to frontier technology forces accelerated development of domestic alternatives that might otherwise not have emerged. The SMIC foundry's achievement of seven-nanometer-class and five-nanometer-class chip production using deep ultraviolet multi-patterning techniques - without access to ASML's extreme ultraviolet systems - is perhaps the most significant technology achievement China has demonstrated under sanctions pressure. It is not a clean triumph: DUV-based production carries higher costs, lower yields, and slower iteration cycles than EUV-based production at TSMC. SMIC's advanced node capacity of approximately 45,000 wafer starts per month in 2025, expanding toward 60,000 in 2026, is a fraction of TSMC's leading-edge throughput. The yield rate of Huawei's Ascend 910C is reportedly between 20 and 40 percent - far below industry norms - reflecting the quality penalty that DUV multi-patterning imposes at advanced nodes.

Yet the political and strategic significance of these achievements exceeds their commercial performance. China has demonstrated that a state-directed semiconductor program, backed by sufficient capital and national will, can achieve capabilities at advanced nodes that Western strategists initially assessed as unattainable without EUV access. The implications for the theory underlying American export control policy are significant: if the controls could not prevent China from reaching seven-nanometer capability, the question of whether they can prevent the next generation - five nanometer, three nanometer - becomes more urgent. In December 2025, Chinese scientists at Jiao Tong University unveiled an all-optical chip described as capable of running advanced AI models at speeds and energy efficiencies exceeding the NVIDIA A100. A team at a Shenzhen laboratory reportedly built a prototype EUV lithography machine - using parts from older ASML systems acquired through secondary markets, assembled by former ASML engineers. Whether that prototype represents a genuine breakthrough or another instance of Beijing's track record of overstating semiconductor milestones remains contested; credible analysts project commercial EUV viability no earlier than 2030 even if the prototype claims are taken at face value.

TSMC and Taiwan: The Geopolitical Fulcrum

Taiwan Semiconductor Manufacturing Company is the single most strategically significant company on earth. It produces the overwhelming majority of the world's most advanced chips, including those used in every device of consequence to American military capability. Its geographic position in Taiwan makes it simultaneously the asset that both the United States and China most need to secure and the potential victim of any military confrontation in the Taiwan Strait. Washington's strategy of constraining China's access to advanced chips while simultaneously constructing alternative foundry capacity in the United States reflects the recognition that TSMC's current concentration in Taiwan is a strategic vulnerability of the first order - one that no amount of export control architecture can fully address.

ASML and the Allied Chokepoint

ASML, the Dutch lithography giant, produces the only EUV machines that can manufacture chips below seven nanometers at commercial scale. Every semiconductor foundry that produces chips for NVIDIA, Apple, AMD, or Qualcomm - every chip that powers American AI development - runs through ASML's machines. The Netherlands' decision to align its export control policy with American restrictions, barring EUV sales to China since 2019, has been the single most consequential allied technology denial measure of the semiconductor war. The durability of that alignment, in the face of Trump administration policy oscillations and the loss of billions of dollars in Chinese revenue for ASML and its Dutch supply chain, is among the most important variables in the long-term trajectory of the contest.

Military and Security Implications

The military dimensions of the semiconductor contest are not secondary to the economic ones - they are primary. The US Department of Defense's assessment, articulated in the 2021 National Security Commission on Artificial Intelligence report, was direct: if China achieves a decisive lead in AI-enabled military systems, it will gain the upper hand in every domain of warfare. That assessment reflects a technical reality: advanced AI applications - autonomous weapons guidance, intelligence fusion, electronic warfare, predictive maintenance, logistics optimization, and cybersecurity - all require continuous access to high-performance computation. An adversary that can train superior AI models on larger datasets, faster, will accumulate compounding advantages across every domain of military competition.

Huawei's Ascend chips, whatever their limitations relative to NVIDIA's H100 or H200, are not toys. The Chinese AI laboratory Zhipu AI trained its GLM-5 model - a 744-billion-parameter system - entirely on Huawei Ascend 910B clusters in 2025, achieving performance on standard benchmarks that approaches Western frontier models. DeepSeek, the Chinese AI startup that shocked global markets in January 2025 with its R1 model - a system that roughly matched the capabilities of GPT-4-class models while requiring dramatically less compute - demonstrated that China's AI research community can achieve results that challenge American assumptions about the relationship between chip access and AI capability. DeepSeek's subsequent optimization of its V4 model for Huawei Ascend hardware was a deliberate strategic statement: China's premier AI laboratories are committing to the domestic hardware stack, validating the Ascend ecosystem with real, frontier-class workloads.

The military application of China's domestic AI chip ecosystem extends beyond model training. The People's Liberation Army's doctrine of intelligentized warfare - articulated across multiple white papers and operational planning documents - envisions AI-enabled autonomous systems, machine-speed decision cycles, and cross-domain coordination that requires exactly the kind of high-performance AI chips that US export controls are designed to deny. China has implemented a military-civil fusion strategy that systematically reduces the barrier between commercial AI chip development and military AI application. When ByteDance or Alibaba deploys Huawei Ascend chips for commercial AI workloads, they are simultaneously advancing the military-grade AI hardware ecosystem through the same infrastructure. The export controls that the Bureau of Industry and Security assessed in May 2025 as having been violated by Huawei's Ascend development reflect precisely this dual-use dynamic.

For the United States military, the semiconductor contest has an equally urgent domestic dimension. The CHIPS Act-funded fabrication facilities in Arizona and Texas represent not merely commercial investments but national security infrastructure - the effort to ensure that the United States can manufacture the chips that its own weapons systems require without depending on a geopolitically exposed Taiwan. The F-35 fighter, submarine systems, missile guidance, and satellite communications all rely on advanced semiconductors. China's 2025-2026 export controls on rare earth magnets - restricting the materials essential to defense electronics manufacturing - represent Beijing's counter-move against American defense supply chain vulnerabilities. The January 2026 implementation of strict licensing requirements for rare earth exports to foreign military end-users, combined with the broader export control framework applied to firms with as little as 0.1 percent Chinese-origin rare earth content, was a direct threat to Western defense production capacity of a sophistication that Washington acknowledged it was not fully prepared to absorb.

Economic and Trade Impact

The economic costs of the semiconductor war are asymmetric, contested, and in many cases self-inflicted on both sides. For American chip companies, the direct revenue impact of China access restrictions has been enormous. NVIDIA's China revenue, historically a major share of its total, collapsed following successive export control rounds. The Information Technology and Innovation Foundation estimated that full US-China chip decoupling could cost American chipmakers 77 billion dollars in lost sales, with downstream effects on the research and development budgets that fund the innovation necessary to maintain the technological leadership that justifies the controls in the first place. AMD, Applied Materials, Lam Research, and KLA all absorbed material China revenue reductions. The paradox is acute: the export controls that are designed to prevent China from developing competitive chips also defund the American research enterprise that is supposed to generate the next generation of American chip leadership.

For China, the economic costs have been differently distributed. The immediate impact of the 2022 controls - price spikes for restricted chip types, workforce reductions at facilities unable to access foreign equipment, and disruption across the semiconductor ecosystem - was significant but not decisive. The longer-term effect has been the redirection of enormous state capital into domestic semiconductor development, producing industrial capacity that Chinese policymakers regard as a strategic asset regardless of its near-term commercial efficiency. China's mature node semiconductor production - chips at 28 nanometers and above, not subject to the most restrictive export controls - has expanded at four times the rate of global demand growth between 2014 and 2025, accounting for approximately half of global mature node capacity. Chinese chipmakers have aggressively designed out American equipment from their production processes, substituting domestic tools from companies including NAURA, AMEC, and Piotech, whose revenues have expanded substantially under the 50 percent domestic equipment mandate that Beijing introduced in 2025.

The rare earth countermeasure deployed by Beijing in 2025 has concentrated minds in Washington and European capitals in ways that chip export controls alone had not. China controls approximately 90 percent of global rare earth processing, 70 percent of mining, and 93 percent of magnet manufacturing - inputs essential to semiconductors, defense electronics, electric vehicles, and a wide range of advanced manufacturing. Beijing's licensing requirements for rare earth exports, effective December 2025, with automatic denial for foreign military end-users, produced export price spikes of up to sixfold for affected materials and licensing approval rates for European firms below 25 percent. The neodymium-praseodymium oxide benchmark surged approximately 40 percent in August 2025 following a single shipment disruption. Washington's initiative - described as Pax Silica - to build alternative critical mineral supply chains with trusted partners in Australia, Canada, and Africa represents the strategic recognition that the chip war is simultaneously a materials war, and that China's leverage in the materials domain is at least as consequential as America's leverage in chip technology.

Diplomatic Positioning

The diplomatic architecture of the semiconductor war involves a set of relationships that are simultaneously the source of American strategic leverage and its greatest vulnerability. The Chip4Alliance framework - coordinating export controls across South Korea, Taiwan, Japan, and the Netherlands - is the practical mechanism through which Washington extended its technology denial strategy beyond what US unilateral controls alone could achieve. Japan introduced restrictions on 23 categories of semiconductor manufacturing equipment in 2023. The Netherlands required ASML to obtain licenses for shipments of older deep ultraviolet systems to China - a restriction that significantly curtailed Chinese chipmakers' access to the equipment they were using to work around EUV denial.

The coherence of that coalition is under pressure from multiple directions simultaneously. Trump's readiness to loosen chip export controls as part of bilateral trade diplomacy with China signals to allied capitals that the United States may be willing to trade away the strategic asset that they have sacrificed commercial relationships to protect. When Beijing agreed to ease rare earth export restrictions in exchange for a relaxation of H20 controls during the Geneva tariff de-escalation talks in 2025, it demonstrated exactly the transactional dynamic that semiconductor policy experts warned against: treating foundational technology controls as bargaining chips in trade negotiations rather than as long-term strategic commitments. Allied governments - whose domestic chip equipment industries have absorbed the commercial cost of the controls - observed this exchange with concern. Dutch and Japanese officials who negotiated their own control regimes in explicit coordination with American commitments face domestic political questions about the durability of those commitments when Washington's own behavior suggests they are negotiable.

China's diplomatic counter-strategy has been to fragment the allied coalition by differentiating between its treatment of individual allied nations. Beijing has pursued bilateral economic engagement with the Netherlands - whose ASML is China's single most critical technology supplier - while simultaneously pressing South Korea and Japan on trade and territorial issues that create friction within their respective relationships with Washington. The removal of Samsung and SK Hynix's named Chinese facilities from the US Validated End-User program, effective December 2025, increased pressure on South Korean chipmakers while raising questions in Seoul about American management of a policy that disproportionately burdens Korean firms in China.

Regional Fallout

Taiwan occupies the most acutely sensitive position in the semiconductor war's regional dimensions. TSMC's fabs represent the largest single concentration of irreplaceable semiconductor manufacturing capacity on earth, and their geographic location in a territory whose political status is the most dangerous flashpoint in US-China relations creates a strategic vulnerability that no amount of CHIPS Act investment in Arizona can fully address on a five-year timeline. TSMC's Arizona fabs, even when fully operational, are expected to produce a fraction of the advanced node capacity that the Hsinchu and Tainan facilities currently deliver. Washington's strategy of simultaneously strengthening Taiwan's semiconductor status as a deterrent and building alternative capacity to reduce vulnerability to a Taiwan contingency is inherently contradictory - and both Beijing and Taipei are aware of that contradiction.

South Korea's position is equally complex. Samsung and SK Hynix are critical players in both the high-bandwidth memory market - where Chinese AI chip programs have significant dependencies - and in advanced foundry capacity. Seoul has aligned with Washington's export control framework while maintaining substantial economic relationships with China. The removal of Samsung and SK Hynix's Chinese facilities from the Validated End-User program created direct commercial pressure on Korean firms that operate major production lines in China. South Korea's semiconductor industry, like Taiwan's, is caught between its alliance commitment to the United States and its commercial exposure to China - a tension that Beijing has consistently sought to exploit.

Japan's semiconductor revival - driven partly by domestic security reassessment and partly by American pressure and incentives - has produced significant investment in domestic chip manufacturing, including a landmark joint venture between the Japanese government and TSMC in Kumamoto. Japan's restrictions on semiconductor manufacturing equipment exports to China have been maintained more consistently than American controls, reflecting a strategic assessment in Tokyo that the technology denial logic is sound even when Washington's implementation is erratic. India, as an emerging semiconductor destination - with incentives attracting both foundry investment and chip packaging capacity - represents the longer-term attempt to build supply chain resilience across a broader allied base.

Global Strategic Consequences

The semiconductor war's most consequential global dimension is its role in driving the bifurcation of the world's technology ecosystem into two distinct spheres. The combination of US export controls, Chinese domestic investment, and the politicization of supply chain relationships is producing a trajectory in which the global semiconductor ecosystem - which operated as a single, highly integrated market for three decades - is fracturing into American-aligned and Chinese-aligned segments. Chinese AI developers are optimizing workloads for Huawei Ascend hardware. Chinese foundries are building domestic equipment supply chains. Chinese chip designers are reducing dependencies on American electronic design automation tools. The short-term cost of this decoupling is enormous - for both sides. The long-term consequence is a world with two parallel technology stacks, each developing along separate innovation trajectories, with compatibility diminishing over time.

The implications for global AI governance are profound. If the United States and China develop distinct AI hardware ecosystems, they will also develop distinct AI model architectures, training paradigms, and application philosophies - shaped in part by the hardware constraints and design choices of each ecosystem. Nations in the Global South, in Southeast Asia, in the Middle East, and in Africa that are building AI infrastructure today will make foundational choices between these ecosystems - choices that carry geopolitical alignment implications extending well beyond the technology domain. Saudi Arabia, the UAE, India, and Indonesia are all actively engaging with both American and Chinese AI infrastructure providers, seeking to maintain optionality precisely because they recognize that their choice of semiconductor ecosystem is also, implicitly, a choice about which great power's technology standards they will embed in their critical systems for a generation.

The DeepSeek episode of January 2025 - when a Chinese AI startup released an open-source model roughly matching GPT-4-class capability while reportedly requiring dramatically less computational power to train - sent a signal that the relationship between chip access and AI capability is less deterministic than American strategy assumed. If Chinese AI researchers can achieve near-frontier results through algorithmic efficiency and model architecture innovation rather than raw compute scale, the export control strategy of denying China computational power may need to be supplemented by controls on AI model weights, training methodologies, and software frameworks that are far more difficult to enforce. The BIS addition of AI model weights to the Commerce Control List in January 2025 represented a recognition of this dynamic - but controlling the diffusion of model weights is a qualitatively different enforcement challenge than controlling the export of physical hardware.

Risk Matrix

• Risk Level: Critical - A Taiwan Strait military contingency triggers the most severe technology decoupling in history, with TSMC fab operations disrupted, American semiconductor supply chains for defense systems immediately stressed, and both the US and China forced to operate on degraded chip production capacity simultaneously.
• Risk Level: Critical - China achieves viable commercial EUV lithography domestically by 2028-2030, rendering the entire architectural premise of Western export controls obsolete and eliminating the manufacturing technology chokepoint that has been the strategy's most durable leverage point.
• Risk Level: High - Trump administration chip policy oscillations fracture the Chip4Alliance framework, with the Netherlands and Japan independently reassessing their commercial exposure to China and allowing controlled equipment exports that undermine the multilateral denial architecture.
• Risk Level: High - China's rare earth export restrictions escalate to a full embargo on critical minerals for Western defense manufacturing, creating acute supply chain failures in F-35 production, missile systems, and submarine electronics that cannot be rapidly remediated.
• Risk Level: High - Chinese AI laboratories achieve, through algorithmic efficiency rather than chip performance, AI capabilities in autonomous weapons and electronic warfare that match or exceed American capabilities despite compute disadvantage - validating the DeepSeek model at military scale.
• Risk Level: Medium - SMIC achieves commercially viable five-nanometer production using DUV multi-patterning by 2027, providing China's AI chip programs with meaningful next-generation capability ahead of Western forecasts and accelerating Huawei Ascend deployment across the PLA.
• Risk Level: Medium - The 15-percent revenue fee mechanism applied to American chip sales to China is expanded to other technology categories, establishing a precedent in which US technology controls function as a permanent tax on American exports to strategic competitors, distorting global technology trade patterns in unpredictable ways.
• Risk Level: Low (near-term) - China achieves genuine semiconductor self-sufficiency across the full value chain - design, equipment, fabrication, packaging - within five years. The structural depth of the Western lead, particularly in EUV lithography and electronic design automation, makes this outcome extremely unlikely before 2032 at the earliest.

Scenario Analysis

Scenario One: Controlled Bifurcation (Most Probable, 12-24 Month Horizon)

The most likely near-term trajectory is a continuation of the current pattern: American chip controls oscillate between tightening and loosening in response to diplomatic and commercial pressures, while China continues to build domestic chip capability incrementally under state direction. The allied export control framework frays at the margins but holds on the most critical chokepoint - EUV lithography - because the Netherlands and Japan recognize that losing ASML's strategic alignment with Washington would mark the definitive failure of the technology denial strategy. China's domestic AI chip market share reaches 50 percent by end of 2026 but Chinese chips remain materially behind American frontier performance on the most demanding training workloads. The world operates with two increasingly distinct but not yet fully separate technology ecosystems, with major developing nations hedging between them. The semiconductor war produces no decisive outcome - it is a grinding attrition contest in which the ultimate victor is the side that sustains its innovation investment, its allied coherence, and its domestic industrial base over the longer arc.

Scenario Two: Diplomatic Détente and Selective Access (Moderate Probability)

The Trump-Xi diplomatic engagement, catalyzed by the May 2026 Beijing summit, produces a framework agreement that trades selective semiconductor access for Chinese concessions on rare earths, fentanyl precursors, and trade balances. The United States institutionalizes a tiered access model - with advanced chips available to approved Chinese commercial customers under license and revenue-sharing, while maintaining absolute denial at the most advanced frontier - effectively monetizing the technology gap rather than closing it. China accepts this arrangement as a pragmatic measure that accelerates its compute capacity while it continues domestic chip development. The allied coalition's coherence degrades as the transactional framing of American chip policy makes multilateral coordination increasingly difficult to sustain. The long-term consequence is that China's AI development accelerates relative to a baseline of complete denial while the revenue mechanism generates US Treasury income that partially offsets CHIPS Act costs.

Scenario Three: Accelerated Decoupling (Lower Probability, Higher Consequence)

A Taiwan crisis, a major sanctions violation resulting in comprehensive technology sanctions against Chinese tech giants, or a congressional override of executive-branch chip export relaxations triggers rapid and comprehensive US-China technology decoupling. TSMC's Taiwan operations face operational uncertainty. Chinese foundries accelerate investment in domestic equipment and EUV alternatives. American chip companies lose access to the Chinese market entirely, forcing painful restructuring while redirecting development resources toward allied markets. The global semiconductor industry bifurcates sharply and irreversibly into two separate innovation ecosystems, each developing independently for the remainder of the decade. AI development in China and the West proceeds on divergent hardware and software architectures, with countries across the Global South forced to make definitive alignment choices. Military AI development on both sides accelerates under resource pressure, producing autonomous systems whose interaction dynamics have not been modeled by strategists on either side.

Intelligence Forecast (6-24 Months)

The six-to-twelve-month horizon will be shaped primarily by the diplomatic legacy of the Trump-Xi Beijing summit of May 2026 and its semiconductor-specific commitments. NVIDIA's H200 sales to approved Chinese customers will proceed - but the absence of actual chip deliveries as of mid-2026, attributed to legal and regulatory limbo under the bilateral tech rivalry framework, means that the practical impact on China's compute capacity will be deferred. Chinese firms led by ByteDance are preparing orders worth up to 14 billion dollars for H200 units; the pace at which those orders are fulfilled will be a leading indicator of whether the diplomatic semiconductor détente is substantive or performative.

On China's domestic capability trajectory, Huawei is targeting production of approximately 600,000 Ascend 910C units in 2026 - roughly twice the prior year's output - with a planned doubling of SMIC's seven-nanometer capacity concurrent. SMIC's five-nanometer pilot runs, targeting mass production for Huawei and Alibaba, will be closely watched; if yield rates improve to commercially sustainable levels, China's advanced node capacity will have achieved a milestone that Western control architects did not anticipate on this timeline. The Ascend 950DT, targeted for launch in the fourth quarter of 2026 with significantly enhanced memory capacity and interconnect bandwidth, will provide a data point on whether Huawei's chip roadmap - which analysts noted was projecting a capability decline in the near term rather than the normal advancement - can recover trajectory.

On the equipment front, ASML's 2026 China revenue decline will be a structural indicator of the allied control framework's durability. Chinese domestic equipment firms' progress in DUV indigenization - particularly the testing of domestic DUV tools by SMIC under its 50 percent domestic equipment mandate - will determine whether China can close the equipment gap through domestic development on a timeline that renders the EUV chokepoint less decisive than Western strategists assume. The Shenzhen EUV prototype, even if commercially unviable before 2030, signals that China's national commitment to closing this gap is unlimited in resource terms.

On the rare earths front, the one-year suspension agreed at the Busan summit of October 2025 will expire, forcing renewed negotiation or escalation on both the chip control and rare earth control dimensions simultaneously. The question of whether Washington can develop sufficient alternative rare earth processing capacity through Pax Silica partnerships to reduce its vulnerability to Chinese export restrictions - and on what timeline - will be a critical determinant of American negotiating leverage in that confrontation.

The twelve-to-twenty-four-month horizon will be shaped by the 2027 tariff policy review cliff and the congressional dynamics of the AI Overwatch Act. If congressional hawks succeed in imposing legislative constraints on executive semiconductor export policy, the transactional flexibility that Trump has employed will be curtailed - potentially restoring greater strategic discipline at the cost of diplomatic friction with Beijing. If the executive branch maintains flexibility, the semiconductor war will continue to be fought simultaneously as a technology denial campaign and a revenue-generating trade management exercise, with strategic clarity sacrificed to commercial and diplomatic pragmatism.

Final Strategic Takeaway

The semiconductor war between the United States and China is a contest without precedent in the history of great-power competition. No previous strategic rivalry has been contested simultaneously at the level of physical manufacturing, intellectual property, software architecture, supply chain geography, alliance diplomacy, and basic scientific research - all at the same time, all in the same industry. The chip is not merely a product. It is the infrastructure of intelligence, the hardware of warfare, the substrate of economic modernity. Whoever leads in semiconductor capability leads in every domain that matters to twenty-first-century power.

The American position is structurally stronger than it appears from the noise of policy oscillation. The US leads in chip design by a generation. ASML's EUV monopoly remains intact. TSMC's advanced node capability is unmatched. The CHIPS Act investments, despite political controversy, are building redundant foundry capacity that reduces strategic vulnerability. The allied framework - however strained - has maintained the EUV chokepoint that is the single most consequential technology denial measure in the war.

The Chinese position is stronger than pure performance metrics suggest. SMIC has achieved seven-nanometer capability without EUV. Huawei's Ascend ecosystem is becoming commercially self-sustaining through state-directed adoption. DeepSeek has demonstrated that Chinese AI researchers can achieve frontier-class capability with constrained compute. China's rare earth leverage is a genuine strategic weapon. And the 47.5-billion-dollar Big Fund III represents a sustained institutional commitment to semiconductor self-sufficiency that shows no sign of wavering under pressure.

The fundamental failure in American strategy is the treatment of foundational technology controls as a trade variable rather than a security commitment. The semiconductor architecture - the chokepoints in design, equipment, and materials - is the most powerful technology leverage the United States has ever possessed in peacetime strategic competition. Using it as a bargaining chip in tariff negotiations, monetizing it through revenue-sharing fees, and treating it as a diplomatic concession in summit diplomacy does not merely weaken its immediate effectiveness. It signals to allies that the commitment is unreliable, to China that it is negotiable, and to the American tech industry that the rules governing its most critical market will shift with each news cycle.

The lesson of the Cold War technology competition is not that export controls alone determine outcomes. It is that sustained, coherent, multilaterally anchored technology denial - combined with domestic investment to widen the lead rather than merely preserve it - can shape the trajectory of a rival's military and economic modernization over decades. The United States won that contest not primarily through denial, but through the superior innovation that the open society generated and the alliance system that extended its technological advantage to partners who shared both its security interests and its scientific capacity. That model is available today. The question is whether Washington possesses the strategic patience and institutional discipline to execute it.

The chip is not the weapon. The chip is the factory that builds every other weapon. Whoever controls silicon controls the future - and the future does not wait for policy coherence to catch up with strategic reality.