Chapter 4: High Technology Competition

Executive Summary

On October 7, 2022, the Biden administration unveiled what industry analysts immediately recognized as the most comprehensive technology export controls in a generation. The new regulations banned sales of advanced semiconductors and chipmaking equipment to China, restricted American personnel from supporting Chinese semiconductor development, and extended controls to foreign-made products using American technology—through the expansive Foreign Direct Product Rule. Nvidia's A100 and H100 AI chips, essential for training cutting-edge artificial intelligence models, could no longer be shipped to Chinese customers. ASML's most advanced extreme ultraviolet (EUV) lithography machines, already blocked from Chinese buyers since 2019, were now joined by deep ultraviolet (DUV) tools capable of producing chips below 14 nanometers. American engineers working for Chinese semiconductor firms faced career-ending choices: resign or lose U.S. export privileges. The United States would sacrifice billions in commercial revenue to prevent China from accessing technologies deemed critical to both economic competitiveness and military superiority.

China's response came swiftly, though not through official statements. Within weeks, Chinese state media declared a "technology self-reliance" campaign, echoing rhetoric from previous embargoes but with renewed urgency and resources. The Big Fund III semiconductor investment vehicle reportedly secured over $47 billion—dwarfing previous efforts (Triolo and Greene 2023). Huawei, whose advanced chip development had been crippled by earlier U.S. restrictions, stunned analysts in August 2023 by releasing the Mate 60 Pro smartphone powered by a sophisticated 7-nanometer chip produced by SMIC (Allen and Weinstein 2023) (Semiconductor Manufacturing International Corporation) using older DUV equipment. While technologically inferior to TSMC's leading-edge 3nm chips and produced with reportedly low yields, the achievement demonstrated Chinese determination and capability to circumvent export controls through indigenous innovation, however costly and inefficient.

The central paradox is inescapable: Washington is trying to strangle a competitor whose supply chains are wrapped around American industry. Unlike Cold War technology denial—where the Soviet Union operated in a separate ecosystem—this competition occurs within deeply integrated supply chains. American semiconductor firms derive 30-40% of revenue from Chinese customers (SIA 2023). Chinese researchers publish more AI papers than their American counterparts. Huawei, Ericsson, Nokia, and Samsung fight over global telecommunications infrastructure based on geopolitical alignment as much as technical merit. Technology competition inflicts costs on the restricting nation even as security concerns override economic logic.

Why does technology define great power rivalry? Because technological capabilities determine both economic competitiveness and military effectiveness. Semiconductors power everything from smartphones to missile guidance systems. AI enables autonomous weapons and facial recognition surveillance. Quantum computing threatens the encryption protecting financial transactions and military communications. The nation that leads gains advantages across every domain—military, economic, political. Technology supremacy has become strategic imperative, not commercial preference.

The dual-use problem makes clean separation impossible. The same AI chips training consumer chatbots also train military targeting systems. CRISPR gene editing treats cancer and could engineer bioweapons. Hypersonic materials enable civilian spacecraft and nuclear delivery vehicles. Commercial technology trade inevitably transfers military capabilities. Governments must choose: economic integration that benefits adversaries, or restrictions that handicap domestic industries while merely delaying adversary development.

But individual technologies matter less than innovation ecosystems. China can purchase AI chips. Building world-class AI requires research universities, venture capital, immigration policies attracting global talent, IP protection, and cultural tolerance for entrepreneurial failure. The United States led technological innovation through ecosystem advantages, not specific policies. No planned economy has achieved sustained technological leadership. Yet American advantages are eroding: Chinese R&D spending now rivals American levels, and Chinese universities produce increasingly competitive research.

Ecosystem Competition: More Than Just Technology Technology competition is ultimately ecosystem competition. Individual technologies can be copied or purchased, but the institutional frameworks that generate sustained innovation—research universities, venture capital, IP protection, immigration policies, entrepreneurial culture—cannot be easily replicated. The fundamental question is whether China's state-directed approach can match America's market-driven ecosystem, or whether authoritarian governance inherently limits innovation capacity. History suggests the latter, but China's scale and determination create genuine uncertainty.

What is at stake is which nation shapes the 21st century. British industrial dominance determined the 19th-century order. American innovation supremacy shaped the 20th. Now we face a starker question: can state-directed development match market-driven innovation? The answer will determine not just U.S.-China relations but global technology governance, alliance structures, and whether supply chains fragment along geopolitical lines.

---

The Semiconductor Battleground

Why Semiconductors Define Technology Competition

Semiconductors occupy a unique position in technology competition: they are simultaneously ubiquitous (powering virtually all modern electronics), strategically critical (essential for both economic activity and military systems), and characterized by extreme concentration in production (with Taiwan's TSMC holding 90% of advanced chip manufacturing). This combination—universality, strategic importance, and geographic concentration—makes semiconductors the single most critical chokepoint in contemporary technology competition.

The strategic significance of semiconductors extends across every domain examined in this book. Military systems depend on advanced chips: F-35 fighter jets contain thousands of chips enabling avionics, sensors, weapons targeting, and communications. Missile guidance systems require radiation-hardened chips operating in extreme conditions. Radar and electronic warfare systems process massive data streams using specialized chips. Naval vessels depend on chips for navigation, combat systems, and command networks. The technological sophistication of modern militaries correlates directly with semiconductor capabilities—a nation lacking access to advanced chips cannot field competitive 21st-century military systems.

Economic competitiveness increasingly relies on semiconductors. Data centers processing cloud computing, AI training, and digital services run on cutting-edge chips—Nvidia's AI accelerators, Intel's Xeon processors, AMD's EPYC chips. Telecommunications networks depend on networking chips from Broadcom, Marvell, and others. Automotive industry transformation toward electric and autonomous vehicles requires chips for battery management, sensors (LIDAR, radar, cameras), driver assistance systems, and entertainment. Consumer electronics—smartphones, laptops, gaming systems, IoT devices—are essentially sophisticated chip delivery mechanisms. The nation or nations controlling semiconductor design and manufacturing hold competitive advantages across virtually every economic sector.

National security vulnerabilities emerge from semiconductor dependencies. A nation relying on foreign chip supplies faces potential cutoffs during crises—imagine U.S. military systems depending on Chinese chips during Taiwan Strait conflict, or Chinese telecommunications dependent on American chips during trade war escalation. This dependency creates what Chapter 1 termed "weaponized interdependence": whoever controls chokepoints can restrict access for strategic advantage. The semiconductor supply chain's complexity—with different stages concentrated in different countries, as examined in Chapter 2—means that multiple nations hold potential choke points: U.S. design tools (EDA software), Dutch lithography equipment (ASML), Taiwanese fabrication (TSMC), American/Japanese materials (photoresists, silicon wafers). Each represents a potential vulnerability if geopolitical alignment fractures.

The semiconductor industry also exhibits characteristics that make competition particularly intense and restrictions particularly potent. Technology leadership requires continuous innovation—the "Moore's Law" dynamic where chip capabilities roughly double every two years means that falling behind technologically creates gaps difficult to close. A semiconductor manufacturer two generations behind (e.g., producing 14nm chips when leaders make 3nm) faces not just quantitative disadvantage but qualitative gaps: lower performance, higher power consumption, larger size, higher costs. These gaps compound: inferior chips mean inferior end products (slower computers, shorter battery life, less AI capability), reducing competitiveness and revenues, limiting R&D investment, widening technological gaps further.

Capital intensity creates natural barriers to entry. Building a cutting-edge semiconductor fabrication facility costs $15-20 billion and requires 3-5 years. A single EUV lithography machine costs $150-200 million, and fabs need dozens. Annual R&D spending for leading firms (TSMC, Samsung, Intel) exceeds $15-20 billion each. This capital intensity means that semiconductor leadership cannot be achieved through incremental investment—it requires sustained, massive commitments over decades. China's semiconductor investment, while enormous in absolute terms ($100+ billion across Big Fund I, II, and III), must compete with private sector investment by Samsung, TSMC, Intel, and others that collectively exceeds this annually. Catching up requires not just matching current investment but exceeding it sufficiently to close gaps while leaders continue advancing.

Supply chain complexity examined in Chapter 2 creates multiple chokepoints across the value chain. The United States dominates chip design (Qualcomm, Nvidia, AMD, Intel, Apple) and essential design tools (Synopsys, Cadence, Mentor Graphics control 100% of EDA software). The Netherlands monopolizes extreme ultraviolet lithography through ASML—literally the only company globally producing EUV machines essential for sub-7nm chips. Japan dominates critical materials (photoresists from JSR/Tokyo Ohka, silicon wafers from Shin-Etsu/SUMCO) and production equipment (Tokyo Electron). Taiwan controls advanced fabrication through TSMC. South Korea provides memory chips (Samsung, SK Hynix). China handles much assembly and testing but controls no critical upstream stages. This distributed supply chain means that comprehensive technology denial requires coordination across multiple countries—or that any single chokepoint can halt production.

October 2022 Export Controls: Strategic Logic and Mechanisms

The Biden administration's October 7, 2022 semiconductor export controls represented a sharp escalation in U.S.-China technology competition and a fundamental shift in export control philosophy. Previous controls targeted specific companies (Huawei, SMIC) or specific applications (military, surveillance). The October 2022 rules instead imposed broad, capability-based restrictions: any semiconductor manufacturing equipment enabling production of chips below 14nm, any advanced AI chip with specified computing capabilities, and any American person supporting Chinese semiconductor development faced restrictions. (Chapter 6 details the Export Administration Regulations (EAR), Entity List designations, and ECCN classification system that provide the legal architecture for these controls.) The breadth shocked industry: controls extended far beyond military applications to encompass nearly all advanced commercial chip development in China.

Strategic logic behind the controls reflected shifting U.S. government assessment of Chinese technology development and security implications. Officials concluded that previous incremental restrictions—Entity List additions targeting specific firms, licenses required for certain sales—were insufficient to prevent Chinese military modernization and potential displacement of American technology leadership. Chinese firms like SMIC, despite Entity List designation, continued advancing (achieving 7nm production demonstrated by Huawei's 2023 chip). Chinese AI development progressed despite limited access to cutting-edge chips, using older but still capable hardware and algorithmic innovations. The U.S. government, facing political pressure to address Chinese technology competition and genuinely concerned about military implications of AI and advanced computing, opted for comprehensive restrictions targeting China's semiconductor ecosystem broadly rather than specific firms.

National Security Advisor Jake Sullivan articulated the strategic shift in a September 2022 speech: moving from maintaining a "relative" advantage (keeping the United States a few years ahead) to establishing "as large a lead as possible" through denying China access to technologies that could close gaps (Sullivan 2022). The old playbook—stay ahead by running faster—gave way to a new one: stay ahead by hobbling the competition. This reflected recognition that dual-use technologies transferred for commercial purposes inevitably support military applications, and that allowing Chinese semiconductor self-sufficiency would empower potential adversary across military and economic domains. Whether this logic succeeds depends on effectiveness across the five criteria examined in Section 1.3, but the strategic intent was clear: accept commercial costs to prevent Chinese technology development.

From Running Faster to Kneecapping the Competition The October 2022 controls represent a fundamental departure from decades of U.S. technology policy. Previously, the strategy was to maintain leads by out-innovating competitors. The new approach explicitly aims to slow adversary progress through denial, accepting that this will also impose significant costs on American firms. This shift acknowledges that in dual-use technologies, commercial sales inevitably transfer military-relevant capabilities.

Mechanisms employed multiple complementary tools to restrict access:

Advanced chip export restrictions prohibited sales of AI chips exceeding specified performance thresholds to Chinese customers. Nvidia's A100 and H100 chips, dominating AI training and inference markets, immediately fell under restrictions. AMD's MI250 accelerators faced similar bans. These restrictions targeted capability rather than specific companies—any chip meeting performance criteria faced controls. Nvidia attempted to develop "China-compliant" chips (A800, H800) with marginally reduced specifications designed to evade controls while maintaining commercial viability. The Commerce Department responded by tightening specifications in October 2023 updates, closing loopholes and banning the compliant variants. This cat-and-mouse game illustrated both industry resistance to lost revenue and government determination to enforce comprehensive restrictions.

Semiconductor manufacturing equipment controls banned sales of equipment capable of producing chips below 14nm to Chinese fabrication facilities. This targeted deep ultraviolet (DUV) lithography tools from ASML (Netherlands), etching and deposition equipment from Lam Research and Applied Materials (United States), and metrology tools from KLA (United States). Previous restrictions had blocked EUV machines (only from ASML) essential for sub-7nm production, but DUV tools could produce 14nm and 7nm chips through multiple patterning—technically challenging but feasible. The October 2022 rules closed this gap, aiming to freeze Chinese semiconductor manufacturing at trailing-edge nodes. ASML faced particular pressure: while its EUV machines were already blocked, its DUV tools (Twinscan NXT) were profitable exports to China. Dutch government implementation of controls faced industry lobbying and political resistance, though ultimately complied following U.S. pressure.

EUV vs. DUV Lithography: Understanding the Technology Gap Extreme Ultraviolet (EUV) lithography uses 13.5nm wavelength light to etch transistor features smaller than viruses, enabling chips at 7nm and below. Deep Ultraviolet (DUV) uses 193nm light and can reach 7nm only through "multi-patterning"—exposing each layer multiple times at enormous cost and complexity. ASML is the sole global manufacturer of EUV machines (each costs $150-200 million). Without EUV, China cannot economically produce cutting-edge chips regardless of how much it invests in other capabilities.

U.S. person restrictions prohibited American citizens and permanent residents from supporting Chinese semiconductor development without Commerce Department authorization. This provision shocked industry: American engineers working for Chinese firms (including SMIC, YMTC, Yangtze Memory Technologies) faced immediate job losses or forced resignations. Senior technologists, many ethnic Chinese who had studied in the United States and worked for American firms before returning to China, found expertise suddenly prohibited. The restrictions targeted human capital transfer—recognizing that equipment alone is insufficient without expertise to operate it effectively. This "brain drain" reversal attempted to halt technology transfer through personnel, though legality faced challenges (restricting Americans' employment based on foreign employer's nationality raises constitutional questions) and effectiveness remains uncertain (non-American engineers could substitute, though with lower expertise).

Foreign Direct Product Rule (FDPR) extension expanded U.S. jurisdiction extraterritorially to cover foreign-made products incorporating American technology. The FDPR, previously applied to Huawei specifically, became general policy: any semiconductor manufacturing equipment made anywhere globally using American technology (software, components, technical data) above de minimis thresholds required U.S. export licenses for sales to Chinese semiconductor fabs. This provision asserted that American technology embedded in foreign products grants U.S. government veto power over sales—a controversial claim of jurisdiction that allied governments privately resented but largely accepted given dependence on U.S. semiconductor technology and markets. The FDPR's effectiveness depends on American content in foreign equipment: if non-U.S. suppliers can substitute American components, the rule's leverage diminishes; if American technology proves irreplaceable, the rule grants comprehensive control.

Extraterritorial Jurisdiction: The Long Arm of American Technology The Foreign Direct Product Rule is an extraordinary assertion of American legal authority over products manufactured entirely outside the United States. If a Dutch lithography machine contains American-origin software or components, Washington claims the right to veto its sale. This extraterritorial reach works only because American technology is often irreplaceable—but it generates resentment among allies and creates incentives for developing non-American alternatives.

Allied Coordination: Success and Tensions

U.S. export controls on semiconductors cannot succeed unilaterally—equipment chokepoints reside in Netherlands (ASML lithography), Japan (Tokyo Electron equipment, JSR photoresists), and South Korea (memory chips, production expertise). Allied cooperation is essential, yet involves tensions between American security imperatives and allied commercial interests.

Netherlands became critical because ASML monopolizes EUV lithography and produces the most advanced DUV tools. The Dutch government initially resisted U.S. pressure for comprehensive DUV restrictions, citing economic costs (China represented ASML's largest market outside Taiwan/South Korea, generating billions in revenue) and legal complexities (Dutch export control law requires independent assessment of military applications, not blanket restrictions based on buyer nationality). Extended negotiations involving U.S. Commerce Secretary Raimondo, Dutch Prime Minister Rutte, and ASML executives eventually produced January 2023 Dutch export control expansion covering advanced DUV tools. However, restrictions proved narrower than U.S. officials desired: only the most advanced DUV machines (capable of producing sub-14nm chips through multiple patterning) faced restrictions, while less advanced DUV tools remained exportable. This reflected Dutch compromise: cooperate sufficiently to maintain U.S. alliance while limiting economic damage to ASML.

Japan faced similar pressures across multiple companies. Tokyo Electron, Japan's leading semiconductor equipment manufacturer, derives substantial revenue from Chinese sales. JSR and Tokyo Ohka supply photoresists critical for lithography—materials where Japanese firms dominate globally. Japanese government implemented export controls in March 2023 covering 23 categories of advanced semiconductor equipment, aligning with U.S. and Dutch restrictions. However, Japanese controls similarly reflected compromise: targeting equipment for leading-edge chips while allowing continued sales of mature-node equipment. Japanese officials emphasized alignment with Wassenaar Arrangement multilateral export control frameworks rather than U.S. unilateral demands, preserving legal distinction even while substantively cooperating.

South Korea presented unique complications. Samsung and SK Hynix, Korea's semiconductor champions, operate fabrication facilities in China producing memory chips (DRAM, NAND flash). U.S. restrictions threatened these operations: if Samsung and SK Hynix couldn't import American equipment or receive support from American engineers, their Chinese fabs would struggle. The U.S. government granted exemptions allowing continued American support for South Korean firms' Chinese operations—exemptions criticized as undermining control effectiveness but deemed necessary to preserve South Korean cooperation. Korean officials faced difficult balance: maintaining alignment with the United States (critical security ally) while protecting Samsung and SK Hynix commercial interests in China (major market).

Effectiveness of allied coordination remains incomplete. While Netherlands, Japan, and South Korea implemented restrictions, gaps exist. Chinese firms can still purchase some advanced equipment, obtain maintenance and upgrades for previously purchased equipment (restrictions don't retroactively block support for installed machines), and potentially source from third-country suppliers (though few alternatives exist for critical tools). Allied governments interpret restrictions narrowly, frustrating U.S. officials seeking comprehensive prohibitions. Commercial lobbying pressures allied governments to minimize restrictions—ASML, Tokyo Electron, and equipment manufacturers argue that lost Chinese revenue reduces R&D budgets for future innovation, potentially undermining long-term competitiveness.

The fundamental tension is that allied governments share U.S. concerns about Chinese military modernization but prioritize commercial interests more highly and resist American extraterritorial jurisdiction. Dutch, Japanese, and Korean perspectives view U.S. technology restrictions as partly motivated by commercial competitiveness (protecting American chip firms against Chinese competition) rather than purely security concerns. This creates suspicion that the United States seeks allied cooperation in restricting Chinese technology access while American firms simultaneously gain market share from European and Asian competitors handicapped by lost Chinese sales.

Future coordination faces challenges. If Chinese technology development continues despite restrictions, allied governments may conclude that accepting economic costs for ineffective restrictions is poor policy. If restrictions prove effective but Chinese retaliation targets allied exports (China restricting rare earth exports to Japan, agricultural purchases from Netherlands, or Samsung market access), allied political support for restrictions could erode. Sustaining coordinated export controls requires balancing security imperatives against commercial pressures—a balance that shifts with political leadership changes, economic conditions, and effectiveness assessment.

Chinese Responses: Adaptation, Circumvention, and Indigenous Development

China's responses to semiconductor export controls operate across multiple dimensions: immediate adaptations to restrictions, attempts to circumvent controls, long-term indigenous development, and potential retaliation. Immediate adaptations involve working within constraints. Chinese AI developers facing restrictions on Nvidia A100/H100 chips shifted to alternative approaches: using older but still capable chips (Nvidia V100, available before restrictions), clustering many weaker chips to approximate fewer powerful chips (less efficient but functional), optimizing algorithms to reduce computing requirements (achieving similar AI performance with less hardware), and purchasing restricted chips through third-country intermediaries or smuggling (explicitly prohibited but difficult to prevent entirely). Baidu, Alibaba, Tencent, and other Chinese tech giants stockpiled AI chips before restrictions took effect, providing buffer capacity for near-term development. These adaptations mean that restrictions slow but don't halt Chinese AI progress—developers adjust to constrained resources rather than abandoning efforts.

Circumvention attempts exploit loopholes and enforcement limitations. Chinese shell companies established in third countries (Singapore, Malaysia, Taiwan) purchase restricted equipment and chips, claiming end-use in permitted locations before diverting to China. Equipment manufacturers' foreign subsidiaries sell to Chinese customers with technical modifications claimed to fall outside control specifications. Individual smugglers purchase chips in small quantities (below regulatory thresholds) and aggregate shipments. U.S. and allied customs enforcement struggles with verification—distinguishing permitted sales (Chinese consumer electronics manufacturing using mature-node chips) from prohibited applications (advanced semiconductor development) requires technical expertise and investigation resources that overwhelm enforcement agencies. Commerce Department investigations and prosecutions send signals but cannot halt all circumvention.

Indigenous development represents China's long-term strategy and the ultimate determinant of restriction effectiveness. Chinese government, recognizing that dependency on foreign technology creates vulnerability, has poured resources into domestic semiconductor capabilities across the value chain:

Semiconductor fabrication investments aim to build SMIC and other Chinese foundries capable of producing advanced chips without foreign equipment. SMIC's achievement of 7nm production demonstrated progress: using older DUV tools and advanced multiple patterning techniques (industry estimates suggest 5+ masks per layer, far more complex than TSMC's EUV process), SMIC produced Huawei's Kirin 9000s chip powering the Mate 60 Pro smartphone released August 2023. This shocked American officials who believed Chinese firms lacked technical capacity for sub-10nm production without EUV machines. However, the achievement comes with caveats: yields variously estimated at 15-40% (compared to TSMC's 90%+ yields), meaning high costs and limited production volumes; process instability requiring extensive trial-and-error; technological ceiling around 5-7nm without EUV (further miniaturization requires extreme ultraviolet lithography China cannot access); and performance gaps versus TSMC's 3nm chips (Huawei's chip less power-efficient and lower-performing than Apple's latest processors using TSMC 3nm).

Huawei Mate 60 Pro: Proof of Concept Under Pressure The August 2023 release of Huawei's Mate 60 Pro smartphone, powered by a domestically-produced 7nm chip, demonstrated that export controls slow but do not halt Chinese technological progress. SMIC achieved this through brute-force multi-patterning using older equipment—technically impressive but economically unsustainable at scale. The achievement showed Chinese determination and capability while also revealing the limits: yields were low, costs were high, and further miniaturization without EUV appears impossible.

Equipment development targets building Chinese alternatives to ASML, Applied Materials, Lam Research, and Tokyo Electron. Shanghai Micro Electronics Equipment (SMEE) is developing DUV lithography tools, but current capabilities lag ASML by 10-15 years (SMEE's best machines comparable to ASML's tools from 2010). AMEC, Naura, and other Chinese equipment firms produce etching, deposition, and metrology tools, but achieve specifications suitable for 14nm+ nodes rather than cutting-edge <7nm. Breakthroughs in EUV lithography—requiring extraordinary precision in optical systems, high-power laser technology, and ultra-clean manufacturing—appear beyond Chinese capabilities for the foreseeable future despite substantial R&D investment. Equipment development is the hardest component: while semiconductor design can potentially achieve innovation through talent and software, equipment manufacturing requires decades of accumulated expertise in precision engineering, materials science, and production processes difficult to replicate quickly.

Materials and chemicals development addresses dependencies on Japanese photoresists, American specialty gases, and other inputs. Chinese firms have achieved progress in some materials (achieving acceptable photoresist quality for mature nodes), but cutting-edge materials remain dependent on Japanese suppliers. This dependence creates additional chokepoints beyond equipment—even if China built indigenous fabrication equipment, reliance on foreign materials leaves vulnerability.

Design tools represent another critical gap. Synopsys, Cadence, and Mentor Graphics (Siemens EDA) control 100% of Electronic Design Automation software essential for designing complex chips. U.S. restrictions prohibit sales of EDA tool updates to Chinese firms, freezing their capabilities at older software versions. Chinese EDA companies exist but produce tools suitable for simple chips, not cutting-edge designs. Building world-class EDA software requires decades of development—accumulated libraries, verification tools, and optimization algorithms cannot be replicated quickly. This dependency means that even if China achieved fabrication independence, design tool limitations would constrain capabilities.

Talent development involves training engineers, physicists, chemists, and technicians with expertise in semiconductor development. China produces more STEM graduates than any country globally and has repatriated many ethnic Chinese engineers who studied and worked abroad. However, leading-edge semiconductor development requires not just quantity but quality: tacit knowledge from operating cutting-edge fabs, experience troubleshooting complex manufacturing processes, and creative problem-solving for unprecedented challenges. U.S. restrictions on American personnel supporting Chinese semiconductor work aim to cut off this expertise transfer, though non-American engineers (from Taiwan, South Korea, Europe) can substitute with varying effectiveness.

Effectiveness Assessment: Five Criteria Analysis

Applying the five effectiveness criteria from Chapter 1 to semiconductor export controls reveals mixed results: controls impose significant costs and delays but face limitations in achieving stated objectives.

Target Compliance: Moderate effectiveness. Chinese firms cannot legally purchase restricted chips and equipment from American, Dutch, or Japanese suppliers, forcing either compliance or circumvention. Major firms (SMIC, Huawei) largely comply with explicit restrictions, avoiding sanctions risk, though aggressive exploitation of loopholes continues. Smaller firms and research institutes pursue circumvention more aggressively. Overall, restrictions force behavioral changes (stockpiling permitted equipment, shifting to indigenous development, purchasing through intermediaries) rather than full compliance.

Capability Degradation: Moderate to high short-term effectiveness, uncertain long-term. Chinese semiconductor capabilities face immediate constraints: lack of EUV machines prevents sub-5nm production, restricted AI chips slow large-scale model training, loss of American engineering expertise hampers troubleshooting complex processes. SMIC's 7nm achievement demonstrates technical capability but economic unviability (low yields, high costs). Chinese firms are 3-5 years behind TSMC/Samsung in fabrication capabilities and lack clear path to close gaps without accessing restricted equipment. However, long-term degradation depends on whether indigenous development succeeds—a question with enormous uncertainty. If China achieves equipment breakthroughs (unlikely but not impossible), degradation proves temporary; if equipment gaps persist (more likely), Chinese semiconductor capabilities remain constrained at trailing-edge nodes indefinitely.

Cost Imposition: High effectiveness. Restrictions impose multiple costs on China: lost revenue for Chinese firms unable to compete in advanced markets, massive R&D investment required for indigenous development (hundreds of billions with uncertain payoff), economic inefficiency from using inferior domestic alternatives (SMIC's low-yield 7nm production far costlier than TSMC), and opportunity costs from engineers and resources devoted to duplicating foreign capabilities rather than advancing beyond them. Chinese semiconductor industry estimates suggest restrictions will reduce Chinese production by tens of billions annually and delay advanced chip availability by years. Whether these costs prove sustainable (China accepts costs to achieve independence) or motivate policy changes remains to be seen.

Sustainability: Moderate concerns. U.S. restrictions require allied cooperation, which faces commercial pressures and political shifts. If Republican or Democratic administrations in allied countries prioritize economic growth over strategic competition with China, export control cooperation could erode. Chinese retaliation targeting allied exports creates political constituencies opposed to restrictions. However, security concerns about Chinese military modernization and technology competition provide countervailing political support. Sustainability also depends on restriction effectiveness: if controls successfully constrain Chinese capabilities, political support for continuing them may strengthen; if Chinese breakthroughs render restrictions ineffective, sustaining them becomes politically difficult.

Collateral Damage: Moderate to high. American semiconductor equipment and chip firms lose tens of billions in annual Chinese revenue—Nvidia alone estimates $7-10 billion in lost AI chip sales (Del Rey 2023). These revenue losses reduce R&D budgets, potentially undermining long-term American innovation leadership. Allied equipment manufacturers (ASML, Tokyo Electron) similarly lose major revenue sources. American semiconductor firms employ tens of thousands of workers whose jobs depend partly on Chinese sales—lost revenue creates political pressure against restrictions. Scientists and researchers face constraints on collaboration with Chinese colleagues, slowing scientific progress. Global supply chain disruption from semiconductor restrictions creates ripple effects across industries dependent on chips. However, security considerations and long-term competitiveness concerns outweigh commercial costs in U.S. government calculations—accepting near-term collateral damage to prevent long-term strategic disadvantage.

Strategic Implications: The Semiconductor Decoupling Dynamic

Semiconductor export controls have initiated a decoupling process with profound strategic implications extending far beyond chips themselves. This decoupling creates several dynamics:

Technology ecosystem fragmentation: Global semiconductor supply chains are splitting along geopolitical lines. Chinese firms increasingly source equipment, materials, and services from Chinese or non-aligned suppliers, building parallel but inferior ecosystem. Western firms consolidate supply chains among allies (U.S.-Japan-Netherlands-South Korea-Taiwan), reducing China exposure. This fragmentation creates inefficiencies (duplicated R&D, smaller economies of scale) but increases resilience (reduced mutual vulnerability).

Allied coordination requirements: Effective semiconductor restrictions require unprecedented allied cooperation on export controls. This cooperation extends beyond semiconductors to other critical technologies (quantum, AI, biotech), creating technology alliance structures paralleling military alliances. The U.S.-Japan-Netherlands semiconductor equipment coordination may presage broader "Tech 10" or similar groupings coordinating technology policies. However, maintaining cooperation faces challenges from diverging commercial interests and political changes.

Chinese determination for self-sufficiency: Export controls have convinced Chinese leadership that dependency on Western technology creates unacceptable vulnerability. This drives enormous investment in indigenous development regardless of economic efficiency. China's semiconductor self-sufficiency pursuit may sacrifice economic growth for strategic autonomy—a calculation Western market economies rarely make. The question is whether China's state-directed approach can achieve technological leadership or whether innovation ecosystems favoring market-driven entrepreneurship retain advantages.

Race to technological leadership: Semiconductor competition catalyzes broader technology race. Both the United States and China are massively increasing R&D spending, industrial policy support, and talent development across semiconductors, AI, quantum computing, and emerging technologies. This spending acceleration could drive innovation beneficial for humanity (medical advances, climate technologies, scientific breakthroughs) or waste resources on duplicated efforts and strategic competition. Historical parallels (Space Race during Cold War drove innovation but at enormous cost) suggest both outcomes are possible.

Military-technological escalation risks: As civilian technology gaps widen between China and the West, Chinese military modernization may plateau, potentially constraining Chinese regional ambitions. Alternatively, constraints could motivate Chinese military investment in asymmetric technologies (hypersonics, AI-powered autonomous systems, cyber capabilities) that don't require cutting-edge semiconductors. U.S. military advantages depend not just on restricting Chinese chip access but on sustaining American innovation—a challenge requiring continued R&D investment, immigration enabling talent acquisition, and education systems producing engineers.

The semiconductor battleground thus represents the sharpest edge of U.S.-China strategic competition, combining economic stakes (hundreds of billions in commerce), security imperatives (military modernization dependencies), and long-term competitiveness (technological leadership determining great power status). How this competition evolves—whether restrictions successfully constrain Chinese capabilities, whether Chinese indigenous development succeeds, whether allied coordination sustains, and whether collateral damage proves acceptable—will shape the 21st-century international order.

---

Artificial Intelligence and Compute - The New Strategic Resource

AI as Dual-Use Technology: From Language Models to Weapons Systems

Artificial intelligence has emerged as the defining technology of the 21st century, with applications spanning from consumer services (ChatGPT, image generation, recommendation algorithms) to military systems (autonomous weapons, target recognition, intelligence analysis). This dual-use character—where the same technology enables both beneficial civilian applications and potentially destabilizing military capabilities—creates profound challenges for export controls and technology competition.

Modern AI, particularly large language models and deep learning systems, depends critically on compute—massive computational power for training and running models. Training GPT-4, Claude, or comparable frontier models requires thousands of specialized AI accelerator chips running for months, consuming tens of millions of dollars in electricity and computing costs. If oil was the strategic resource of the 20th century, compute may be the strategic resource of the 21st. Access to advanced AI chips determines which nations, companies, and research institutions can develop cutting-edge AI. Unlike previous technologies where expertise and algorithms dominated, contemporary AI transforms compute into a bottleneck—those controlling AI chip supply can influence global AI development trajectories.

Compute as Strategic Resource: The New Oil AI development has transformed raw computational power into a strategic resource analogous to oil in the 20th century. Training a frontier AI model requires thousands of specialized chips running for months at costs exceeding $100 million. Only a handful of institutions globally can afford this, creating a new form of technological inequality. Nations lacking access to advanced AI chips face constraints on AI development that will increasingly determine economic and military competitiveness.

Military AI applications demonstrate why governments treat AI as strategic technology requiring restriction. Autonomous weapons systems—drones that identify and engage targets without human intervention—depend on AI vision systems and decision-making algorithms trained on vast datasets using powerful chips. China has publicly showcased drone swarms capable of coordinated action, reportedly using AI algorithms developed by Chinese tech firms. Target recognition systems enabling missiles to distinguish military from civilian targets, evade countermeasures, and adapt to battlefield conditions require AI trained on extensive imagery and sensor data. Intelligence analysis leveraging AI to process satellite imagery, communications intercepts, and open-source data enables militaries to identify patterns, predict adversary actions, and optimize resource allocation. Cyber operations increasingly employ AI to identify vulnerabilities, craft phishing campaigns, and automate intrusions at scale beyond human operators' capacity.

The U.S. military's Project Maven, initiated in 2017, exemplifies military AI adoption: using machine learning to analyze drone footage (Shane and Wakabayashi 2018), identifying objects and patterns faster and more accurately than human analysts. The project sparked controversy when Google employees protested company participation, ultimately leading Google to withdraw—illustrating tensions between commercial AI development and military applications. However, other firms (Palantir, Anduril, Shield AI, and increasingly Microsoft, Amazon, and Oracle) embrace defense AI contracts, recognizing strategic importance and commercial opportunities. China faces no similar corporate resistance: civil-military fusion doctrine explicitly mandates private sector support for military AI development, with Baidu, Alibaba, Tencent, SenseTime, and others actively collaborating with People's Liberation Army on AI systems.

Surveillance and social control applications demonstrate AI's political implications. China's surveillance state leverages AI facial recognition (SenseTime, Megvii, Hikvision systems) to monitor populations, track dissidents, and enforce social control in Xinjiang and beyond (Mozur 2019). These systems depend on AI chips for real-time processing of millions of video streams, matching faces against databases, and identifying "suspicious" behaviors. Western governments condemn such applications while developing their own surveillance AI (for border control, counterterrorism, and law enforcement), creating hypocrisies and dilemmas about AI governance. Export controls on AI chips partly aim to prevent empowering authoritarian surveillance, though effectiveness faces limitations: older chips (A100's predecessor V100) retain substantial surveillance capabilities, and algorithms continue improving efficiency even with constrained hardware.

Economic and scientific applications explain why AI restrictions face commercial resistance and potential long-term costs. AI powers drug discovery (predicting molecular interactions, identifying potential treatments), climate modeling (simulating atmospheric dynamics with unprecedented resolution), materials science (discovering new compounds for batteries, catalysts, semiconductors), and financial markets (algorithmic trading, risk assessment, fraud detection). Restricting AI chip exports to China means cutting Chinese researchers off from frontier AI capabilities—potentially slowing scientific collaboration and reducing global innovation even while protecting American leads. American pharmaceutical companies collaborating with Chinese research institutions face disruptions. Climate research requiring Chinese participation (China operates major climate models and represents critical data source) encounters obstacles. These costs don't disappear simply because they serve long-term strategic interests—they represent genuine tradeoffs requiring justification.

The October 2022 AI Chip Restrictions: Logic and Implementation

AI chip restrictions announced alongside semiconductor controls (Section 1.2) targeted Chinese access to computing power required for frontier AI development. The logic was straightforward: if training cutting-edge AI models requires thousands of high-performance chips, restricting access to those chips constrains Chinese AI capabilities, maintaining American advantages in both military and commercial AI applications.

Specific restrictions focused on chip capabilities rather than applications. The October 2022 rules prohibited exports to China of chips exceeding specified thresholds of computing performance and chip-to-chip interconnect bandwidth:

• Compute density threshold: 600 trillion operations per second (TOPS) for INT8 operations and 300 teraflops (TFLOPS) for FP16/FP32

• Interconnect bandwidth threshold: 600 gigabytes per second (GB/s)

These specifications targeted Nvidia's flagship AI chips: the A100 (released 2020) providing 1,200 GB/s interconnect bandwidth and 624 teraflops FP16 performance, and H100 (released 2022) with even higher specifications. AMD's MI250 accelerators faced similar restrictions. The thresholds aimed to prohibit chips optimal for large-scale AI training while potentially permitting less capable chips for inference (running trained models) and other applications.

Nvidia's response demonstrated commercial creativity in circumventing restrictions while technically complying. Within weeks, Nvidia announced "China-compliant" chips: the A800 (modified A100) and H800 (modified H100) with interconnect bandwidth reduced to 400 GB/s—just below the 600 GB/s threshold—while maintaining computing performance. These chips sacrificed some efficiency in distributed training (where multiple chips must exchange data rapidly) but retained substantial capabilities for AI development. Nvidia could legally export these variants, preserving billions in Chinese revenue while ostensibly respecting control requirements.

The Commerce Department's October 2023 update closed these loopholes, revising specifications to ban the compliant variants. New rules set lower thresholds and employed "total processing performance" metrics incorporating both compute density and interconnect bandwidth, preventing specification gaming. Nvidia's ability to sell AI chips to China was effectively eliminated, forcing Chinese customers to either stockpile older chips (V100, T4), use clustered weaker chips with performance penalties, or develop indigenous alternatives (discussed in Section 2.4).

Allied coordination for AI chip restrictions faces less complexity than semiconductor equipment controls because AI chips primarily originate from U.S. companies (Nvidia, AMD, Intel). However, ensuring Chinese customers don't obtain chips through third-country intermediaries requires export compliance and customs enforcement. Reports of Nvidia chips reaching China through Singapore and Hong Kong shell companies, third-country procurement agents, and individual smuggling highlight enforcement challenges. The United States pressured Singapore and Hong Kong to tighten export controls and investigate suspicious transactions, with mixed success. Systematic circumvention proves difficult at scale (Chinese AI firms need thousands or tens of thousands of chips, not ones or tens smuggled individually), but marginal circumvention continues.

AI Research Leadership: Publications, Patents, and Talent

Figure 4.2: AI research output and impact by country, showing publications, citations, and talent distribution.

AI competition extends beyond chips to research capabilities: which countries publish cutting-edge research, file foundational patents, attract and develop top talent, and translate research into commercial products and military applications. Metrics reveal complex dynamics where the United States and China lead in different dimensions while Europe falls behind.

Research publications show China overtaking the United States in quantity while competing in quality. According to Stanford's AI Index and analysis of Scopus publication databases, China publishes more AI/ML research papers annually than the United States (approximately 40-45% of global AI publications versus 10-15% for the U.S.). However, citation metrics—measuring research influence—show American papers cited more frequently on average, suggesting higher impact. The most-cited individual papers and breakthrough publications disproportionately originate from American institutions (Stanford, MIT, Berkeley, OpenAI, Google Research, Microsoft Research) rather than Chinese universities and firms. This pattern suggests Chinese research emphasizes quantity and incremental advances while American research produces more foundational breakthroughs—though China's quality is improving rapidly, with Tsinghua, Peking University, and firms like Baidu and Alibaba publishing increasingly influential work.

Patent filings in AI-related technologies show similar patterns. WIPO data indicates Chinese entities file more AI patents than American counterparts, particularly in applications (computer vision, natural language processing, recommendation systems). American firms file more foundational algorithm and architecture patents. Patent quality assessments (citations, legal scope, commercial value) generally favor American patents, though measuring quality proves methodologically challenging and controversial. Patent filings also reveal different focuses: Chinese patents emphasize applications (surveillance, e-commerce, social media), American patents cover broader algorithmic innovations and AI chips/hardware.

Talent development and circulation represent critical dimensions where the United States has historically held advantages but faces erosion. American universities dominate global AI education and research: Stanford, MIT, Carnegie Mellon, Berkeley, and others train both American and international students (including many Chinese) who disproportionately contribute to AI advances. Immigration historically allowed the United States to retain top foreign talent—Chinese, Indian, European researchers staying in America after PhDs to join Google, Microsoft, OpenAI, or start companies. This "brain gain" amplified American AI capabilities beyond domestic population proportions.

Recent dynamics threaten American advantages. Increasingly, Chinese AI researchers return to China after U.S. education, attracted by generous compensation, research funding, and opportunities to lead large teams. Alibaba, Baidu, Tencent, ByteDance, and other firms offer competitive salaries and access to massive datasets (hundreds of millions of Chinese users) unavailable to Western researchers. Chinese government funding for AI research rivals or exceeds American levels (though measuring precisely proves difficult due to definitional differences and reporting opacity). Geopolitical tensions, visa restrictions following Trump administration policies, and COVID-19 pandemic disruptions reduced Chinese student numbers in U.S. universities—potentially cutting future American access to Chinese AI talent.

The United States also develops domestic talent, though shortfalls exist. Computer science enrollments at American universities have surged, but demand for AI expertise exceeds supply, driving fierce corporate competition for talent (six-figure salaries for new PhDs, million-dollar compensation for experienced researchers). Many top American students pursue lucrative industry careers rather than academic research, potentially reducing long-term foundational research. Educational outcomes show the United States producing fewer STEM graduates per capita than China, South Korea, and several European nations, though American education quality at top institutions remains unmatched globally.

Transfer between research and application represents another dimension where American ecosystem advantages manifest. Silicon Valley's venture capital, startup culture, and major tech firms excel at commercializing AI research: OpenAI transformed GPT research into ChatGPT commercial product within months, Google commercialized transformer architectures into search and advertising improvements, and countless startups leverage academic research for specialized applications. China's AI commercialization proceeds rapidly (TikTok's recommendation algorithms, Alibaba's logistics optimization, SenseTime's surveillance systems), but ecosystem differences create advantages for American translation of research to products, particularly in global markets where Chinese firms face regulatory and political obstacles (TikTok bans, Huawei restrictions, data localization requirements limiting Chinese cloud services).

Chinese AI Development: Constrained but Not Halted

U.S. restrictions on AI chips aim to constrain Chinese AI development, yet Chinese capabilities continue advancing through adaptations, stockpiling, algorithmic innovations, and indigenous chip development. Assessing whether restrictions achieve meaningful delays or merely inconvenience Chinese developers requires examining specific Chinese responses and their effectiveness.

Stockpiling and hoarding before and immediately after October 2022 restrictions allowed Chinese firms to accumulate substantial AI chip inventories. Alibaba, Tencent, Baidu, ByteDance, and others reportedly purchased tens of thousands of A100 and H100 chips before restrictions took effect, spending billions to secure near-term supply. These stockpiles enable continued frontier AI development for 2-3 years, during which time Chinese firms can train large models, deploy AI systems, and potentially develop indigenous alternatives. Stockpiling success demonstrates export control challenges: restrictions announced with insufficient lead time (industry knew restrictions were coming months before implementation) allow targets to adapt.

Algorithmic innovations reduce compute requirements, allowing Chinese researchers to achieve competitive AI performance with constrained hardware. Techniques include:

• Model compression and quantization: Reducing model precision from 32-bit to 8-bit or even 4-bit operations dramatically cuts computing requirements with minimal accuracy loss for many applications.

• Efficient architectures: Developing models requiring fewer parameters and operations to achieve equivalent performance—e.g., China's GLM-130B model claims performance comparable to OpenAI's GPT-3 (175 billion parameters) while using only 130 billion parameters.

• Training optimizations: Techniques like mixed-precision training, gradient checkpointing, and efficient parallelization reduce chip requirements for given model sizes.

• Specialized models: Rather than pursuing general-purpose foundation models like GPT-4, developing specialized models for specific applications (language translation, image recognition, recommendation systems) that require less compute.

These innovations mean Chinese AI development continues even with chip constraints—though potentially falling behind American frontier models that leverage unrestricted access to thousands of H100 chips. Chinese researchers publish extensively on efficient AI, potentially sharing advances globally and benefiting non-Chinese researchers—an irony where restrictions motivate innovation that diffuses beyond the target.

Indigenous AI chip development pursues Chinese alternatives to Nvidia and AMD accelerators. Huawei's Ascend 910 chip, developed before U.S. restrictions intensified, provides AI training capabilities comparable to Nvidia's V100 (pre-A100 generation). Cambricon, a Chinese AI chip startup, produces training and inference accelerators used by Chinese firms and government agencies. These chips lag Nvidia's latest generations (Ascend 910 performs roughly at 2018-2019 Nvidia levels), and production faces challenges without access to TSMC or Samsung cutting-edge fabrication (Huawei's Ascend uses 7nm process, now constrained by semiconductor equipment restrictions). However, domestic AI chips partially substitute for embargoed Nvidia hardware, reducing dependence and providing fallback capacity.

Cloud computing workarounds allow Chinese users to access foreign AI computing resources indirectly. Chinese researchers and firms can rent AI compute from foreign cloud providers (Amazon Web Services, Google Cloud, Microsoft Azure, Oracle Cloud) operating in third countries (Singapore, Japan, Europe), ostensibly for permitted purposes while potentially using capacity for restricted AI development. Cloud providers face export compliance obligations and implement customer screening, but distinguishing permitted from prohibited AI development proves challenging. Unless the United States extends restrictions to prohibit cloud AI services to Chinese customers—an escalation with major commercial costs and enforcement complexity—cloud access provides partial circumvention route.

Chinese AI development trajectories remain robust despite restrictions, though likely slowed relative to counterfactual without controls. Chinese firms continue releasing competitive AI products: Baidu's ERNIE models (competitor to ChatGPT), Alibaba's Tongyi Qianwen language model, SenseTime's multimodal AI systems, and TikTok's recommendation algorithms (arguably world-leading in engagement optimization). Chinese military AI development proceeds across autonomous systems, intelligence analysis, and cyber applications. Complete Chinese AI development stagnation proves implausible; whether the United States maintains meaningful leads (2-3 years, 5 years, or permanent) remains uncertain and depends on sustained American innovation, effective enforcement of restrictions, and Chinese indigenous capabilities.

Compute as Strategic Resource: Implications and Alternatives

The centralization of AI capability in advanced compute resources creates a new form of strategic dependency analogous to oil dependence in the 20th century. Nations lacking indigenous AI chip production or access to foreign chips face constraints on AI development, potentially falling behind economically and militarily. This dependency creates several dynamics:

Compute inequalities between nations and institutions shape AI development globally. Only a handful of institutions can afford frontier AI model training: OpenAI (Microsoft-backed), Google DeepMind, Meta, Anthropic (Amazon-backed), and a few others with access to tens of thousands of H100-equivalent chips and tens of millions in compute budgets. Most nations lack resources for comparable efforts—creating AI capabilities dominated by American firms and potentially China. This concentration raises governance questions: should AI development this consequential occur in private firms pursuing commercial objectives rather than democratic oversight? Concerns about AI risks (from misinformation to potential existential threats) intersect with debates about compute access and control.

Cloud concentration amplifies dependencies. Amazon Web Services, Microsoft Azure, and Google Cloud dominate global cloud computing, including AI-specific services providing access to GPU clusters, pre-trained models, and development tools. Firms and researchers globally depend on these platforms for AI development—creating surveillance opportunities (cloud providers can observe model training and usage patterns), control mechanisms (providers can restrict access or raise prices), and leverage for American foreign policy (government can pressure providers to cut off disfavored customers). China's pursuit of indigenous cloud infrastructure (Alibaba Cloud, Tencent Cloud, Huawei Cloud) partly reflects determination to escape this dependency, creating parallel ecosystems.

Energy consumption for AI training and inference creates environmental costs and practical constraints. Training large models consumes tens of gigawatt-hours of electricity—equivalent to hundreds of American households' annual consumption. Global AI expansion could consume substantial percentages of electricity production, raising costs and environmental concerns (if powered by fossil fuels) or requiring renewable energy expansion. Compute-constrained nations may face energy constraints limiting AI development even if they access chips—an underappreciated bottleneck in AI competition.

Alternative AI paradigms less dependent on massive compute could shift competitive dynamics. Neuromorphic computing, inspired by brain architectures, potentially enables AI with far lower energy consumption. Analog computing approaches promise specialized AI chips more efficient than digital alternatives. Quantum-inspired algorithms could achieve specific AI tasks more efficiently than conventional approaches. Breakthroughs in any of these alternatives could render current compute-centric restrictions obsolete—creating both opportunities (for countries innovating successfully) and risks (for those invested in conventional paradigms). China's substantial investment in alternative computing approaches reflects recognition that conventional paths may remain constrained by foreign dependencies.

Diffusion of AI capabilities complicates control strategies. As AI techniques improve, yesterday's frontier capabilities become accessible to broader audiences using less powerful chips. GPT-2, once restricted from release due to misuse concerns, runs on consumer hardware. Image generation models operate on smartphones. This diffusion means that even if the United States successfully restricts Chinese access to cutting-edge chips, previous-generation capabilities—still formidable—diffuse globally. Permanent AI advantage through export controls appears implausible; the goal becomes maintaining leads of 2-5 years rather than permanent dominance.

The AI compute competition is a race where both the United States and China are sprinting, with U.S. restrictions aiming to slow Chinese progress while America maintains innovation pace. Success requires not just restricting Chinese chip access but sustaining American AI innovation through research investment, talent development, and commercial ecosystems supporting rapid deployment. Failure scenarios include both Chinese breakthroughs closing gaps despite restrictions and American complacency assuming restrictions alone suffice without continued innovation.

---

Emerging Technologies - Quantum, Space, and Biotechnology

Quantum Computing and Communications: The Next Frontier

Quantum technologies represent potentially revolutionary capabilities across computing, communications, and sensing—with implications for cryptography, drug discovery, materials science, and military systems. Unlike semiconductors and AI where current capabilities are well-established, quantum technologies remain largely in research and early development phases, making competition about future potential rather than present applications. This uncertainty creates both opportunities (countries achieving breakthroughs could leapfrog competitors) and risks (massive investments may yield limited practical returns if technical barriers prove insurmountable).

Quantum computing leverages quantum mechanical phenomena (superposition and entanglement) to perform certain calculations exponentially faster than classical computers. If scaled successfully, quantum computers could break current encryption systems (threatening financial transactions, military communications, diplomatic secrets), accelerate drug discovery (simulating molecular interactions beyond classical computing capacity), optimize complex systems (logistics, supply chains, financial portfolios), and advance materials science (discovering new catalysts, battery chemistries, superconductors). These applications explain why governments treat quantum computing as strategic priority requiring substantial R&D investment and potential export restrictions.

Current quantum computing capabilities remain limited. The most advanced quantum computers—Google's Sycamore, IBM's Quantum System One, and competitors—contain hundreds of qubits (quantum bits) but face challenges with error rates, coherence times (how long qubits maintain quantum states), and connectivity between qubits. These limitations mean that quantum computers cannot yet outperform classical computers for practical applications—the much-publicized "quantum supremacy" demonstrations involve contrived problems designed to favor quantum approaches rather than real-world tasks. Industry estimates suggest that practical quantum advantage for meaningful applications (beyond demonstrations) may require 1,000-10,000 logical qubits with error correction—a threshold potentially decades away.

U.S.-China quantum computing competition shows both countries investing heavily while following different strategies. American quantum computing efforts combine government funding (National Quantum Initiative allocating billions for research), academic research (MIT, Caltech, Berkeley, Chicago leading), and private sector innovation (Google, IBM, Microsoft, Amazon, and startups like Rigetti, IonQ, PsiQuantum). This distributed ecosystem leverages America's traditional advantages in fundamental research and commercial innovation. China's approach emphasizes concentrated state direction: large government investments in quantum research institutes, dedicated quantum computing centers, and mandates for commercial participation. Chinese researchers publish extensively on quantum computing, and institutions like University of Science and Technology of China (USTC) contribute frontier research.

Publication metrics show U.S.-China rough parity in quantum computing research. Both countries contribute 20-25% of global quantum computing publications, with Europe producing comparable numbers. Citation analysis suggests American and Chinese research compete in quality, with both producing influential papers and neither demonstrating clear dominance. This parity contrasts with AI (where American research leads in citations) and semiconductors (where American design dominates)—quantum computing remains sufficiently early-stage that research leadership is contested.

Quantum communications represents a nearer-term application with important security implications. Quantum key distribution (QKD) leverages quantum mechanics to generate encryption keys with theoretically unbreakable security—any eavesdropping attempt disturbs quantum states and reveals intrusion. China has aggressively deployed quantum communication infrastructure: the Beijing-Shanghai quantum communication trunk line (2,000+ km of fiber optic cable with QKD), the Micius quantum satellite enabling quantum key exchange between ground stations separated by thousands of kilometers, and integration of quantum communications into government and financial networks.

Chinese quantum communication leadership reflects strategic priorities and advantages. The technology requires expensive infrastructure investment (specialized fiber networks, satellite systems) where state-directed funding excels. Applications primarily serve government communications and financial security—domains where Chinese state control enables rapid deployment without requiring commercial viability. Whether quantum communications provide genuine security advantages over post-quantum cryptography (classical encryption algorithms resistant to quantum computing attacks) remains debated among cryptographers, but Chinese deployments demonstrate technical capability and willingness to invest in speculative technologies.

U.S. quantum communication efforts proceed more cautiously, reflecting different priorities and skepticism about quantum key distribution's practical advantages. American researchers argue that post-quantum cryptography—upgrading classical encryption algorithms to be quantum-resistant—may provide equivalent security at lower cost and greater flexibility than quantum key distribution networks. Government funding supports quantum communication research, but deployment lags Chinese efforts. This divergence creates an interesting dynamic: China leads in quantum communication infrastructure deployment, but whether this leadership translates to strategic advantages depends on future quantum computing trajectories and cryptographic developments.

Export controls on quantum technologies remain limited but expanding. The October 2022 semiconductor export controls included restrictions on quantum computing equipment and materials, though specific applications remain classified. Quantum computers require cryogenic cooling systems, specialized control electronics, and precision manufacturing—potential chokepoints where U.S. or allied firms dominate. However, quantum technology's early stage and distributed research ecosystem complicate export controls: much relevant knowledge resides in published research accessible globally, making restrictions on equipment only partially effective. Future export controls likely will target specific quantum computing capabilities (number of qubits, error rates, coherence times) as technology matures and military applications clarify.

Space Systems: Dual-Use Infrastructure and Military Competition

Space capabilities have become essential infrastructure for modern economies and militaries: satellites provide communications, navigation (GPS/Galileo/BeiDou), earth observation (weather, agriculture, intelligence), and increasingly commercial services (internet connectivity, remote sensing). This dual-use character—where civilian and military space applications share common technologies—makes space a domain of intensifying strategic competition with economic and security dimensions.

Launch capabilities determine access to space and represent a critical technology where multiple nations compete. The United States historically dominated space launch through NASA and later commercial firms (SpaceX, Blue Origin, United Launch Alliance). SpaceX's reusable Falcon 9 rockets dramatically reduced launch costs (from $10,000-20,000 per kilogram to $2,000-3,000), enabling satellite constellation deployment and maintaining American launch dominance despite reduced government space budgets. China has rapidly expanded launch capabilities: conducting 60+ orbital launches annually (comparable to or exceeding U.S. totals), developing diverse launch vehicles, and achieving technological milestones (first landing on far side of the moon, Mars rover, space station construction). Russia retains launch capabilities but faces constraints from sanctions and reduced commercial demand. Europe (Ariane), Japan (H-IIA), and India (PSLV/GSLV) maintain independent launch capabilities serving commercial and government needs.

Launch competition combines commercial and strategic dimensions. Commercial launch services represent a market worth billions annually, with SpaceX dominating Western markets and China providing cost-competitive launches for developing nations' satellites (particularly Belt and Road Initiative participants). Strategic competition involves assured access to space: nations dependent on foreign launches face potential restrictions during crises. China's pursuit of independent launch capabilities reflects determination to escape dependency on Western launch services—ensuring that geopolitical tensions cannot cut off space access. U.S. restrictions on satellite exports to China (instituted in 1998 after alleged technology transfer concerns) accelerated Chinese indigenous satellite development, demonstrating how export controls motivate target self-sufficiency.

Satellite constellations for communications and earth observation represent another competitive domain. SpaceX's Starlink constellation (5,000+ satellites operational, plans for tens of thousands) aims to provide global internet coverage, with substantial implications for connectivity in remote and underserved areas. China's state-owned enterprises are developing competing constellations (Guowang/GW with projected 13,000 satellites) serving similar functions. European OneWeb, Amazon's Project Kuiper, and others pursue constellation strategies. These mega-constellations create both opportunities (global internet coverage, bridging digital divides) and concerns (space debris, light pollution, military applications enabling communications in denied areas, surveillance implications).

Navigation systems demonstrate strategic imperatives for indigenous capabilities. GPS (U.S.), GLONASS (Russia), Galileo (Europe), and BeiDou (China) provide positioning and timing services essential for civilian and military applications. Military systems depend on navigation for precision weapons, troop movements, and platform coordination—dependency on foreign navigation systems creates vulnerability if access is restricted during conflicts. China's BeiDou development, completed in 2020 with global coverage, eliminated dependency on GPS and provides services throughout Belt and Road Initiative nations. European Galileo development similarly reflected determination to escape GPS dependency. These parallel systems create redundancy beneficial for civilian users (multiple systems improve accuracy and reliability) while serving strategic autonomy objectives.

Anti-satellite (ASAT) capabilities and space weaponization represent the most troubling military space competition dimensions. China, Russia, and the United States have all demonstrated ASAT capabilities through kinetic interceptors (missiles destroying satellites), directed energy weapons (lasers), electronic warfare (jamming satellite communications), and cyber attacks. China's 2007 ASAT test destroyed a weather satellite, creating thousands of debris pieces that will threaten space operations for decades. Russia has tested co-orbital ASAT systems (satellites maneuvering near targets before attacking). The United States maintains ASAT capabilities while advocating for space norms prohibiting debris-generating tests. This competition creates escalation risks: space assets are vulnerable, attacks are difficult to attribute, and debris from conflicts threatens all space operations indiscriminately.

Space domain awareness and tracking capabilities determine which nations can monitor space activities, detect potential threats, and attribute attacks. The United States maintains the most comprehensive space surveillance network, tracking tens of thousands of objects and providing collision warnings to satellite operators globally. China is developing comparable capabilities through ground-based radars, optical telescopes, and space-based surveillance satellites. This tracking capability serves both peaceful purposes (debris avoidance, space traffic management) and military objectives (targeting adversary satellites, protecting indigenous assets).

Export controls and space technology face significant challenges. International Traffic in Arms Regulations (ITAR) restrict many space technologies as munitions, requiring licenses for exports and limiting international collaboration. These restrictions aim to prevent technology transfer enabling adversary space and missile capabilities but impose costs: American satellite manufacturers lost market share to European competitors less constrained by export controls, and scientific collaboration faces bureaucratic obstacles. Recent reforms have eased some restrictions while maintaining controls on sensitive technologies, but balancing commercial interests against security concerns remains contentious.

Biotechnology and Synthetic Biology: From CRISPR to Biosecurity

Biotechnology represents perhaps the most consequential and least-governed domain of U.S.-China technology competition. Advances in gene editing (CRISPR), synthetic biology (engineering organisms for specific functions), and computational biology (AI-driven drug discovery and protein design) promise revolutionary medical treatments, agricultural improvements, and industrial applications—while raising existential biosecurity risks from engineered pathogens, genetic discrimination, and ecosystem disruption.

CRISPR and gene editing capabilities have rapidly diffused globally since the technology's development in 2012. The United States leads in foundational research (Jennifer Doudna and Emmanuelle Charpentier's Nobel-winning work at Berkeley and European institutions), but Chinese researchers aggressively apply CRISPR to human embryos, agricultural crops, and medical treatments with less ethical and regulatory constraint than Western counterparts. He Jiankui's 2018 creation of gene-edited human babies (attempting to confer HIV resistance) violated international norms and Chinese regulations, resulting in his imprisonment, but demonstrated Chinese willingness to push ethical boundaries in pursuit of technological firsts.

Research leadership in biotechnology shows complex dynamics. American institutions (NIH, major research universities, pharmaceutical companies) publish the most influential biological research, file the most biotech patents, and translate research to commercial products most effectively. China has rapidly increased biotech research output and quality: Chinese institutions now contribute 20%+ of global biological research publications, increasingly in top-tier journals. However, citation patterns and breakthrough discoveries still favor American research. The broader pattern resembles AI: China pursues rapid scaling of research quantity and practical applications while the United States maintains advantages in foundational discoveries and quality.

BIOSECURE Act and restrictions on Chinese biotech firms represent U.S. efforts to limit Chinese biotech access and address data security concerns. The BIOSECURE Act, enacted in December 2025 as part of the FY2026 NDAA, prohibits federal agencies from contracting with designated "biotechnology companies of concern" (BCCs). The enacted law uses the DoD's Section 1260H list rather than naming specific companies — a narrowing from the original House bill that had explicitly targeted WuXi AppTec, WuXi Biologics, BGI Group, and MGI Tech. The stated rationale combines data security concerns (Chinese firms processing American patients' genomic data could transfer information to Chinese government), supply chain vulnerabilities (U.S. pharmaceutical development depending on Chinese contract research organizations creates dependencies), and long-term competitiveness concerns (Chinese firms using American data and collaborations to advance while Chinese data remains inaccessible to Americans).

WuXi AppTec and contract research organizations exemplify Chinese biotechnology's global integration and strategic concerns. WuXi provides drug discovery and development services to Western pharmaceutical companies, handling chemistry, manufacturing, and testing—essentially outsourced R&D. Western firms benefited from cost savings and Chinese expertise, while WuXi gained access to cutting-edge research, established commercial relationships, and developed capabilities. However, this integration creates vulnerabilities: if geopolitical tensions sever relationships, Western pharmaceutical development faces disruptions from lost capacity, while China retains intellectual property and capabilities developed through collaborations. The BIOSECURE Act attempts to force decoupling before dependencies deepen further.

BGI Group and genomic data raises distinct concerns. BGI operates sequencing services globally, including prenatal testing for hundreds of thousands of international customers. Genomic data from these services potentially flows to Chinese databases, creating privacy concerns and strategic intelligence opportunities (population-level genetic data could inform biological weapon development targeting specific ethnicities or identifying genetic advantages for human enhancement). While evidence of malicious data use is limited, the potential for abuse motivates restrictions. BGI's response—arguing that data is anonymized, stored according to local regulations, and serves legitimate commercial purposes—highlights tensions between beneficial international scientific collaboration and legitimate security concerns.

Synthetic biology and dual-use research present governance challenges exceeding those of previous technologies. Synthetic biology enables engineering organisms to produce pharmaceuticals (insulin, vaccines), industrial chemicals (biofuels, materials), and agricultural improvements (nitrogen-fixing crops, pest resistance). The same techniques could engineer pathogens with enhanced transmissibility, lethality, or resistance to treatments—whether for bioweapons development or gain-of-function research investigating pandemic threats. The 2020 COVID-19 pandemic, originating in Wuhan and involving virology research at the Wuhan Institute of Virology, intensified debates about laboratory biosafety, gain-of-function research, and potential for accidental or deliberate pathogen releases.

Export controls on biotechnology remain limited compared to semiconductors or AI, reflecting biotechnology's distributed nature and dual-use civilian applications. Biological weapons convention prohibits offensive bioweapons development, but verification challenges and defensive research exceptions create loopholes. The Australia Group coordinates export controls on biological materials and equipment among 42 countries, but China is not a member. U.S. export controls cover specific pathogens, toxins, and specialized equipment (fermenters, freeze dryers, spray dryers) but cannot restrict widely available knowledge or commercial biological materials. The challenge is that much biotechnology requires only published knowledge, commercial equipment, and standard laboratory practices—making export controls far less effective than for capital-intensive technologies like semiconductors.

Chinese biotechnology strategy emphasizes scaling and commercialization: building large genomic databases (collecting data from hundreds of millions of citizens), supporting biotech firms through subsidies and government contracts, investing in agricultural biotechnology, and pursuing medical breakthroughs (gene therapy, stem cell treatments) with less regulatory constraint than the United States or Europe. This strategy leverages China's population size (enabling massive clinical studies), state-directed funding, and willingness to accept risks Western regulators prohibit. Whether Chinese biotechnology achieves leadership depends on whether scaling advantages and risk tolerance outweigh American advantages in fundamental research quality, intellectual property protection, and regulatory frameworks ensuring safety and efficacy.

Biosecurity risks from U.S.-China technology competition include reduced international collaboration (hampering pandemic preparedness and scientific progress), intelligence collection concerns motivating data restrictions (limiting beneficial research uses), and potential biotech decoupling (fragmenting global supply chains for pharmaceuticals and medical equipment). Unlike semiconductors where military applications are clear, biotechnology's primary applications are civilian (medicine, agriculture, environmental remediation), making restrictions costlier and ethically more fraught. Balancing open science traditions with security concerns represents an ongoing challenge without clear resolution.

---

Innovation Ecosystems and Industrial Policy

R&D Spending and the Race for Investment

Research and development investment determines long-term technological leadership, yet measuring and comparing R&D spending across countries involves definitional challenges, data quality concerns, and questions about efficiency versus quantity. Nevertheless, broad patterns reveal intensifying competition where Chinese R&D investment rivals or exceeds American levels while questions persist about which system generates more innovation per dollar invested.

Figure 4.1: R&D spending by country from 1990-2024, showing the dramatic rise of Chinese investment.

Global R&D trends show dramatic growth and shifting geographic distribution. World Bank and OECD data indicate global R&D spending reached approximately $2.5 trillion annually (roughly 2.5% of global GDP), with growth concentrated in Asia. The United States remains the largest single-country investor at $700-750 billion annually (roughly 3.5% of U.S. GDP), but China has closed gaps dramatically: from less than 10% of U.S. R&D spending in 2000 to 85-90% by 2020, with estimates suggesting Chinese R&D spending may have surpassed American levels by 2023 in purchasing power parity terms. European Union collectively spends comparable amounts to the United States ($400-450 billion), but fragmentation across 27 member states reduces coordination and scale. Japan, South Korea, Taiwan, and other advanced economies maintain high R&D intensity (3-4% of GDP) but represent smaller absolute totals.

U.S. R&D composition exhibits distinctive characteristics reflecting market-driven innovation ecosystems. Private sector funding dominates, accounting for roughly 65-70% of total R&D, with government contributing 20-25% (primarily through NIH, NSF, DOD, DOE), and universities/nonprofits 5-10%. This composition means American R&D responds primarily to commercial incentives and market demands rather than state direction. Major companies (Amazon, Alphabet, Microsoft, Meta, Apple, pharmaceutical firms, aerospace manufacturers) invest tens of billions annually in R&D pursuing competitive advantages. This commercial focus creates efficiencies (market discipline eliminates unproductive research) but potentially underinvests in basic research with long payoff horizons (where commercial returns are uncertain but societal benefits potentially enormous).

Chinese R&D composition reverses the public-private balance: government and state-owned enterprises account for 60-70% of R&D funding, with private sector contributing the remainder. This reflects China's state-directed innovation system where government identifies strategic priorities (semiconductors, AI, quantum computing, aerospace, biotechnology) and directs resources accordingly. Major SOEs (State Grid, PetroChina, Sinopec) conduct substantial R&D as required by government mandates. Private firms (Huawei, Tencent, Alibaba, BYD) invest heavily in R&D but often in coordination with state objectives. This composition enables rapid scaling of investment in priority areas but risks inefficiency from political interference, misallocation to unproductive projects, and lack of market discipline.

Measuring R&D effectiveness proves challenging. Patent filings, publication counts, and citation metrics provide partial indicators but don't capture commercial impact or societal value. American R&D arguably demonstrates higher efficiency: U.S. firms commercialize innovations rapidly (Google's search algorithms, Apple's iPhone, Moderna's mRNA vaccines), translate research to profitable products, and achieve global market leadership across numerous sectors. Chinese R&D shows increasing effectiveness (Huawei's 5G leadership, BYD's electric vehicles, TikTok's algorithm) but faces questions about whether massive investment quantities compensate for lower per-dollar productivity. The fundamental question is whether China's state-directed approach can match or exceed market-driven American innovation—a question with ideological, empirical, and political dimensions where evidence remains mixed.

Universities, Research Institutions, and Talent Development

Research universities represent critical nodes in innovation ecosystems, producing both fundamental discoveries and trained talent that feeds industry R&D. University quality, academic freedom, and connections to industry differ significantly across the United States and China, creating competitive advantages and vulnerabilities.

American universities dominate global rankings of research institutions: Stanford, MIT, Harvard, Berkeley, Caltech, Carnegie Mellon, Princeton, and others consistently rank among the world's top universities for science and engineering. This dominance reflects sustained investment (endowments worth billions, federal research funding exceeding $40 billion annually through NIH, NSF, DOD), academic freedom enabling pursuit of curiosity-driven research, and cultural prestige attracting global talent. American universities train not just American students but international cohorts: 50%+ of STEM PhDs at top programs come from abroad (particularly China and India), creating brain gain when these graduates stay in the United States for industry careers.

Chinese universities have improved dramatically in global rankings, with Tsinghua, Peking University, Zhejiang, Fudan, and Shanghai Jiao Tong now appearing in top-100 global rankings. Chinese government investment in elite universities (Double First Class University Plan allocating tens of billions) aims to match Western institutions. Publication output from Chinese universities has surged, particularly in engineering and applied sciences. However, academic freedom constraints, political requirements (Communist Party cells in universities, restrictions on sensitive research topics), and faculty compensation structures (favoring administrative positions over research) create challenges. Brain drain persists: many top Chinese students educated at Chinese universities pursue PhDs abroad and remain in the West, though returnee rates have increased.

Research institutions beyond universities contribute significantly to national innovation. American national laboratories (Lawrence Livermore, Los Alamos, Sandia, Argonne, Oak Ridge, and others) conduct basic and applied research in areas with limited commercial incentives (nuclear weapons, fundamental physics, advanced materials). DARPA (Defense Advanced Research Projects Agency) funds high-risk, high-reward research that has produced breakthroughs including the internet, GPS, and mRNA vaccine platform technologies. These institutions bridge basic science and applications, taking risks commercial firms avoid.

China's research institutes include Chinese Academy of Sciences (CAS, with hundreds of member institutions), specialized defense research institutes, and newer entities like West Lake University (established 2018, modeled on Western research universities). These institutions receive substantial government funding and pursue state-directed priorities. Whether they can match DARPA's record of breakthrough innovations remains uncertain—success requires tolerance for failure and risk-taking that may conflict with political accountability demands in authoritarian systems.

Talent circulation and "brain drain/brain gain" dynamics critically shape technology competition. The United States has historically benefited from global talent flows: recruiting foreign students to American universities, retaining many through employment visas and green cards, and absorbing scientists and engineers fleeing authoritarian regimes (Soviet scientists after Cold War, Chinese researchers after Tiananmen). This talent importation amplified American innovation beyond what domestic population and education systems alone could produce.

Recent trends create concerns about eroding American advantages. Chinese STEM PhD holders increasingly return to China (estimates suggest returnee rates have risen from 25% in 2000s to 50%+ in 2020s) as Chinese opportunities improve and American visa policies tighten. Trump administration policies (restrictive H-1B visas, security screening for Chinese students, investigations of Chinese researchers under "China Initiative") created hostile environments that pushed talented individuals away. COVID-19 pandemic travel restrictions disrupted international education flows. If these trends continue, American access to global talent—a historic competitive advantage—could diminish while China benefits from reverse brain drain.

Venture Capital, Startups, and Commercialization

Innovation ecosystems require not just research but also mechanisms translating discoveries to commercial products. Venture capital funding, entrepreneurial culture, and regulatory environments supporting startups differentiate American and Chinese innovation systems with implications for technology competition.

Silicon Valley and the broader American venture capital ecosystem represent distinctive competitive advantages difficult for competitors to replicate. Approximately $200-300 billion in venture capital deploys annually in the United States, concentrated in software, biotechnology, and hardware sectors. This capital supports tens of thousands of startups pursuing risky innovations with uncertain payoffs. Most fail (90% of startups eventually shut down or remain small), but successes (Google, Amazon, Facebook, Tesla, Nvidia) generate enormous economic value and technological breakthroughs. The ecosystem combines multiple elements: wealthy individuals and institutions willing to invest in high-risk ventures, experienced entrepreneurs who achieved prior successes and fund/mentor new startups, legal and financial infrastructure supporting equity financing, and cultural acceptance of failure (entrepreneurs who failed previously can raise capital for subsequent ventures).

Chinese venture capital has grown explosively, rivaling American levels with $100-150 billion deploying annually (though estimates vary widely due to definitional differences and data quality). Chinese VC funded successes include Alibaba, Tencent, ByteDance, Meituan, and others worth hundreds of billions collectively. However, Chinese VC differs structurally: state-guided funds (central and provincial government venture funds) account for larger shares than in the United States, influencing investment toward government priorities rather than purely commercial returns. Regulatory uncertainty creates risks: sudden government crackdowns on sectors (private education, gaming, fintech) have destroyed billions in value and discouraged investment. IPO markets favor state-owned enterprises and firms with political connections, creating exit challenges for independent startups.

Entrepreneurial culture differences shape innovation patterns. American culture celebrates entrepreneurship: founders like Elon Musk, Mark Zuckerberg, and others achieve celebrity status, inspiring imitators. Failure carries limited stigma—serial entrepreneurs who failed previously raise capital for new ventures. University students aspire to startup careers as much as corporate employment. This culture supports high-risk innovation: entrepreneurs pursue improbable visions (SpaceX reusable rockets, Moderna mRNA vaccines, OpenAI frontier AI) that established firms avoid.

Chinese entrepreneurial culture has developed rapidly but differs in character. Successful entrepreneurs (Jack Ma, Pony Ma, Zhang Yiming) inspire admiration, but government tolerance for independently powerful entrepreneurs has declined (Jack Ma's Ant Financial IPO blocked, regulatory crackdowns on tech platforms, entrepreneurs required to demonstrate political loyalty). Risk tolerance favors incremental innovations and business model innovations (e.g., applying Western platforms to Chinese markets) over technological breakthroughs. State-owned enterprises dominate sectors deemed strategic, limiting entrepreneurial opportunities. Whether China can sustain rapid innovation with increasing state control remains uncertain.

Regulation and intellectual property protection create ecosystem differences. American regulatory environments generally favor innovation: FDA approves drugs efficiently (relative to alternatives), SEC regulations support startup IPOs, patent system protects intellectual property (imperfectly but better than most countries). This enables startups to raise capital, attract talent, and commercialize innovations with reasonable regulatory predictability.

China's regulatory environment presents greater uncertainties. Patent protection has improved but enforcement remains weaker than the United States—technology theft and counterfeiting persist. Regulatory approvals often favor SOEs and connected firms. Government interventions can suddenly destroy company value (gaming time restrictions harming Tencent, education company bans). These uncertainties create risks that discourage long-term investment and favor established firms over disruptive startups.

Industrial Policy: CHIPS Act vs. Made in China 2025

Government industrial policy—direct state intervention to support specific industries or technologies—has returned to favor after decades of market-oriented skepticism. Both the United States and China now pursue aggressive industrial policies, though with different mechanisms, scales, and philosophies.

The CHIPS and Science Act (2022) represents American industrial policy's most significant recent example, allocating $52 billion in subsidies for domestic semiconductor manufacturing and R&D. The Act aims to reduce American dependence on Asian semiconductor fabrication (particularly TSMC's Taiwan concentration), strengthen supply chain resilience, and maintain technological leadership. Subsidies fund construction of new fabs (Intel, TSMC, Samsung building facilities in Arizona, Ohio, Texas, and elsewhere) and support R&D consortia developing next-generation technologies. Additional tax credits (25% investment tax credit) amplify incentives. The Act represents philosophical shift: accepting that market forces alone won't maintain semiconductor capabilities and that government subsidies, despite inefficiencies, serve national security imperatives.

The Inflation Reduction Act (2022) similarly provides $370 billion in tax credits and subsidies for clean energy technologies: battery manufacturing, solar panel production, electric vehicle assembly, and critical mineral processing. Like CHIPS, IRA reflects industrial policy logic: subsidizing domestic capacity to reduce Chinese dependencies, addressing climate change, and creating manufacturing jobs. Requirements for domestic content and restricted Chinese content (batteries with Chinese materials increasingly ineligible for tax credits) aim to reshore supply chains.

Made in China 2025 represents China's systematic industrial policy across ten strategic sectors: semiconductors, AI, robotics, aerospace, electric vehicles, biotechnology, new materials, agricultural machinery, rail equipment, and maritime engineering. The policy set ambitious targets: 70% domestic content in core components by 2025, global leadership in advanced manufacturing. Implementation involves subsidies (Big Fund I, II, III for semiconductors totaling $100+ billion), preferential procurement (government agencies required to buy domestic products), forced technology transfer (foreign firms accessing Chinese markets must partner with Chinese companies and transfer technology), and R&D support.

Made in China 2025's explicit articulation sparked Western backlash, particularly American concerns about Chinese industrial policy undermining fair competition, stealing intellectual property, and threatening Western technological leadership. The Trump administration's trade war and Biden administration's export controls partly aimed to counter Made in China 2025. Chinese government subsequently downplayed explicit references but pursues substantive policies largely unchanged.

Effectiveness and sustainability of these industrial policies remains contested. American critics argue that government "picking winners" wastes resources on politically favored projects rather than economically viable ventures, that subsidies enrich corporations without generating lasting capabilities, and that protectionism invites retaliation harming exports. Supporters counter that market failures (private sector underinvests in strategic capabilities with national security externalities) justify intervention, and that competing with China's state-directed economy requires government support.

Chinese industrial policy faces different critiques: subsidies encourage overcapacity and inefficiency (steel, solar panels, and potentially semiconductors produced beyond market demand), political priorities override economic logic (supporting SOEs rather than more efficient private firms), and top-down direction misses market signals guiding investment. However, Chinese successes (electric vehicles, high-speed rail, telecommunications equipment) demonstrate that state-directed development can achieve substantial capabilities even if economically inefficient.

The broader pattern is industrial policy convergence: both the United States and China now actively support strategic industries through subsidies, procurement preferences, and trade restrictions. This convergence creates escalation risks (subsidy competitions waste resources) while potentially benefiting technological development (increased R&D spending, manufacturing capacity, innovation incentives). Whether market-driven or state-directed approaches prove more effective likely depends on sectoral specifics, implementation quality, and sustained political commitment rather than ideological superiority.

---

Chinese Perspective Box: Technology Sovereignty and the "Century of Humiliation"

Historical Context: Foreign Technology Denial as National Trauma

Chinese strategic thinking about technology is fundamentally shaped by the Century of Humiliation (百年国耻, bǎinián guóchǐ)—the period from the 1840s Opium Wars through 1949 when technological backwardness enabled foreign domination. British gunboats defeated Qing dynasty forces because superior Western industrial technology produced weapons China could not match. Japanese invasion during World War II exploited this same vulnerability: modern Japanese industry produced aircraft, ships, and armaments that technologically backward China could not counter. For Chinese elites, this historical experience carries a clear lesson: nations lacking indigenous technological capabilities face exploitation, occupation, and destruction of sovereignty.

The early People's Republic reinforced these imperatives. CoCom restrictions denied China access to advanced manufacturing equipment, computers, and dual-use technologies. Soviet technical assistance during the 1950s proved temporary—the Sino-Soviet split in 1960 saw Soviet advisors withdrawn overnight, crippling industrial programs dependent on foreign inputs. China's response was enforced self-reliance. The "Two Bombs, One Satellite" program (两弹一星, liǎngdàn yīxīng) developed nuclear weapons, ballistic missiles, and satellites despite Western embargo and Soviet abandonment, demonstrating that determined state effort could achieve technological breakthroughs from positions of profound backwardness.

Key Chinese Terms in Technology Strategy

Contemporary Chinese technology policy employs specific vocabulary reflecting historical experience and current strategic priorities:

Technological Self-Reliance and Self-Strengthening (科技自立自强, kējì zìlì zìqiáng) represents the core strategic imperative: China must develop indigenous capabilities in critical technologies to eliminate foreign leverage. This term, elevated to policy prominence under Xi Jinping, emphasizes that self-reliance is not merely economic preference but national survival requirement.

Stranglehold/Chokepoint (卡脖子, qiǎ bózi, literally "strangling the neck") describes foreign control over critical technology inputs that could be weaponized against China. When Huawei lost access to Google services and TSMC fabrication, Chinese commentators framed this as Americans "strangling China's neck." The term resonates deeply with Century of Humiliation memories of foreign powers controlling China's destiny through superior technology.

Domestic Substitution (国产替代, guóchǎn tìdài) refers to replacing imported technologies with domestically-produced alternatives. This policy priority drives massive investment in indigenous semiconductors, operating systems, EDA software, and manufacturing equipment—accepting higher costs and lower quality as acceptable prices for eliminating chokepoint vulnerabilities.

October 2022 Controls as Technological Containment

Chinese officials and commentators characterize the October 2022 semiconductor export controls as technological containment (技术遏制, jìshù èzhì)—a deliberate American strategy to prevent China from achieving technological parity. Official statements frame the controls as revealing Washington's true intentions: not fair competition but permanent Chinese subordination.

Chinese Foreign Ministry spokesman Wang Wenbin declared the controls "typical of American technological hegemony" (技术霸权, jìshù bàquán), arguing they violated market principles and international trade norms. State media portrayed the restrictions as confirming that Western integration rhetoric masked containment: China was encouraged to specialize in low-value manufacturing while remaining dependent on Western technology, creating permanent vulnerability to coercion.

The controls validated Chinese arguments for self-reliance. If the United States would restrict access to commercial technologies (AI chips for consumer applications, not military systems), then no level of engagement could guarantee access—only indigenous capability could ensure security.

Made in China 2025 and Semiconductor Self-Sufficiency

Made in China 2025 (中国制造2025, Zhōngguó Zhìzào 2025) established explicit self-sufficiency targets: 40% domestic content in core components by 2020, 70% by 2025. Semiconductors received priority given their foundational role across military and civilian applications.

Semiconductor self-sufficiency goals proved most challenging. China's 2015 starting position—producing roughly 15% of semiconductors consumed domestically, nearly all at mature nodes—required building an entire ecosystem: design, fabrication, equipment, materials, and packaging. The policy recognized that semiconductor dependencies represented existential vulnerabilities: a nation dependent on foreign chips for military systems, telecommunications, and industrial control faces potential paralysis if access is denied.

Western backlash to Made in China 2025 led Chinese officials to de-emphasize explicit references, but substantive policies continued unchanged under different labels.

Big Fund Investments and State-Directed Innovation

The National Integrated Circuit Industry Investment Fund (国家集成电路产业投资基金, commonly called "Big Fund" or 大基金) exemplifies state-directed technology development:

• Big Fund I (2014): $21 billion supporting SMIC, Hua Hong, Yangtze Memory Technologies Corporation (YMTC), and packaging firms

• Big Fund II (2019): $29 billion focusing on design tools (EDA), equipment manufacturing, and materials

• Big Fund III (2024): Estimated $47 billion targeting advanced chips, AI processors, and domestic equipment

These investments reflect Chinese willingness to accept inefficiency for sovereignty. Even if domestically-produced chips cost 30-50% more than imports, eliminating foreign chokepoints justifies the premium. Learning-by-doing builds expertise and human capital with long-term payoffs beyond immediate products.

Big Fund corruption scandals (multiple executives arrested for embezzlement) highlight governance challenges in state-directed investment, but have not diminished commitment to the model. Chinese leadership views market-driven approaches as inadequate for technologies where established Western firms dominate and commercial incentives favor continued offshoring.

Civil-Military Fusion Doctrine

Civil-Military Fusion (军民融合, jūnmín rónghé) mandates integration between civilian technology development and military applications. Under Xi Jinping, this became national strategy requiring private technology firms to support military modernization, share research relevant to defense applications, and accept People's Liberation Army involvement in ostensibly civilian projects.

The doctrine reflects recognition that dual-use technologies—semiconductors, AI, quantum computing, biotechnology—advance both economic competitiveness and military capabilities simultaneously. Separating civilian from military applications is artificial: the same chips powering smartphones enable missile guidance systems.

Civil-military fusion creates complications. Western firms and researchers increasingly avoid Chinese collaboration, international partnerships face restrictions, and Chinese firms dependent on foreign markets must balance government mandates against regulatory risks abroad. American export controls explicitly target civil-military fusion, treating all Chinese technology development as potentially military-relevant.

Dual Circulation Strategy

Post-2020, China's Dual Circulation (双循环, shuāng xúnhuán) strategy restructured economic priorities in response to technology coercion risks:

Internal Circulation (国内大循环, guónèi dà xúnhuán) emphasizes domestic consumption and production, building complete supply chains for strategic sectors (semiconductors, pharmaceuticals, aerospace) within China. This reduces vulnerability to foreign pressure but requires accepting efficiency losses from smaller scale and reduced specialization.

External Circulation (国际循环, guójì xúnhuán) maintains international engagement but from a position of strength rather than dependence. China continues participating in global trade and investment, but strategic sectors must have domestic alternatives ensuring that external disruptions cannot paralyze critical industries.

Dual circulation directly responds to U.S. weaponization of interdependence. If integration creates vulnerability, China must rebalance toward domestic self-sufficiency while maintaining beneficial external relationships where risks are manageable.

Implications for Understanding Chinese Responses

Chinese perspectives on technology competition create predictable dynamics: export controls trigger indigenous development acceleration rather than policy concessions; restrictions validate self-reliance arguments, mobilizing resources and political will; retaliation targets Western vulnerabilities selectively to impose reciprocal costs; and long-term competition focuses on building parallel technology ecosystems rather than accommodation.

Western policymakers expecting Chinese compliance underestimate historical determination rooted in Century of Humiliation memories. Chinese leadership views technological independence as existential—the difference between sovereignty and subjugation. Economic efficiency arguments fail to engage this frame. Understanding these perspectives enables more effective strategy: anticipating responses, avoiding counterproductive escalation, and recognizing that technology competition will persist regardless of policy choices.

---

Case Study: U.S. Semiconductor Export Controls (October 2022-2024)

Background and Strategic Context

The October 7, 2022 semiconductor export controls represented the culmination of escalating U.S.-China technology competition and a fundamental shift in American export control philosophy from targeted entity-specific restrictions to comprehensive capability-based controls. Understanding this shift requires tracing the evolution from earlier incremental measures to the sweeping October 2022 rules and subsequent updates.

Prior to October 2022, U.S. semiconductor restrictions targeted specific Chinese firms deemed national security threats. Huawei's Entity List designation in May 2019 restricted sales of American semiconductors and equipment, expanded in May 2020 to cover foreign-made chips using U.S. technology (Foreign Direct Product Rule). SMIC joined the Entity List in December 2020 due to military ties. These targeted restrictions achieved narrow objectives (degrading Huawei's smartphone and 5G equipment competitiveness, constraining SMIC's most advanced production) but left most Chinese semiconductor industry accessible to American chips and equipment. Chinese AI firms purchased Nvidia GPUs without restriction, Chinese fabs bought ASML DUV tools freely, and American engineers worked for Chinese semiconductor companies legally.

The strategic logic for comprehensive controls emerged from several assessments. First, targeted restrictions proved inadequate: SMIC continued advancing despite Entity List status, achieving 7nm production demonstrated later. Chinese firms not on Entity Lists developed AI capabilities using unrestricted chips. Second, dual-use technologies meant that commercial sales inevitably supported military applications through civil-military fusion. Third, U.S. officials concluded that preventing Chinese semiconductor self-sufficiency served American interests more than maintaining commercial semiconductor sales revenue. Fourth, growing Congressional pressure and bipartisan consensus on China competition created political momentum for aggressive action.

The October 2022 controls imposed three complementary restrictions: (1) advanced AI chip export prohibitions targeting compute capabilities and interconnect bandwidths, (2) semiconductor manufacturing equipment export restrictions targeting tools capable of producing sub-14nm chips, and (3) U.S. person prohibitions banning American citizens and permanent residents from supporting Chinese advanced semiconductor development without licenses. These rules aimed to prevent China from developing cutting-edge chips regardless of which specific firms attempted development.

Analytical Framework Application

Domain: Technology (semiconductors and artificial intelligence) with secondary trade implications.

Target:

• Sectoral: Chinese semiconductor manufacturing industry broadly (all fabs, equipment suppliers, design firms)

• Entity-level: Specific restrictions on firms like SMIC, Huawei, and others on Entity List with enhanced controls

• Individual: U.S. persons restricted from supporting Chinese semiconductor work

Objective:

• Primary: Capability degradation—preventing Chinese development of advanced semiconductors and AI chips

• Secondary: Containment—slowing Chinese technology advancement to maintain U.S. competitive advantages

• Tertiary: Compellence—pressuring China to modify technology development practices and civil-military fusion policies (though unstated and unlikely to succeed)

Intensity: Level 4-5 (Severe to Comprehensive)

• Denies entire categories of technology to adversary regardless of end-use

• Extraterritorial application through FDPR

• Requires allied coordination to achieve effectiveness

• Imposes substantial economic costs on American firms ($7-10 billion annually in lost Nvidia AI chip sales alone)

• Creates technological bifurcation with long-term economic and strategic implications

Implementation and Allied Coordination

Implementation required addressing multiple technical and diplomatic challenges. Commerce Department Bureau of Industry and Security (BIS) published 139 pages of regulations defining controlled capabilities, establishing licensing procedures, and providing compliance guidance. Definitions of "advanced computing chips" evolved as Nvidia and AMD developed "China-compliant" variants (A800, H800) that technically met specifications while maintaining substantial capabilities. October 2023 updates tightened definitions, closing loopholes and banning compliant chips.

Allied coordination proved essential but challenging. Netherlands controls ASML lithography equipment, Japan produces critical materials and equipment (Tokyo Electron, JSR), and South Korea's Samsung and SK Hynix operate fabs in China. Extended negotiations produced Dutch export restrictions (January 2023) covering advanced DUV tools, Japanese controls (March 2023) on 23 equipment categories, and South Korean exemptions allowing continued American support for Korean firms' Chinese operations.

Enforcement mechanisms include export licensing requirements, customs inspections, end-use verification, and penalties for violations. However, circumvention through third-country procurement, smuggling, and cloud computing access continues. Commerce Department lacks resources for comprehensive enforcement, and distinguishing prohibited from permitted transactions requires technical expertise not readily available.

Effectiveness Assessment: Five Criteria

Target Compliance (Moderate): Major Chinese firms largely comply with explicit restrictions while aggressively exploiting loopholes. Stockpiling before implementation, purchasing through intermediaries, and using cloud computing access provide partial circumvention. Smaller firms and research institutes pursue more aggressive circumvention. Overall, restrictions force behavioral changes but not complete compliance.

Capability Degradation (High Short-Term, Uncertain Long-Term): Chinese access to cutting-edge chips and manufacturing equipment is substantially constrained. SMIC's 7nm achievement demonstrates capability but at low yields and high costs unsuitable for commercial scale. China currently cannot produce sub-7nm chips or access Nvidia H100-equivalent AI accelerators. However, long-term degradation depends on whether Chinese indigenous development succeeds—a question with 5-10 year time horizons and enormous uncertainty.

Cost Imposition (Very High): Restrictions impose multiple costs: lost commercial revenue ($50-100 billion annually across Chinese semiconductor industry), massive R&D investment required for indigenous alternatives (Big Fund III $47+ billion), economic inefficiency from using inferior domestic chips, and opportunity costs from resources devoted to duplicating foreign capabilities rather than advancing beyond them. Whether China sustains these costs depends on political determination rather than economic rationality.

Sustainability (Moderate Concerns): U.S. restrictions require allied cooperation subject to commercial pressures and political changes. If allied governments prioritize economic growth over strategic competition, coordination could erode. Chinese retaliation targeting allied exports creates political constituencies opposing restrictions. However, security concerns provide countervailing support. Sustainability also depends on effectiveness: if restrictions successfully constrain China, political support strengthens; if Chinese breakthroughs render restrictions ineffective, sustaining them becomes difficult.

Collateral Damage (High): American firms lose tens of billions in Chinese revenue (Nvidia, AMD, Applied Materials, Lam Research, KLA). Revenue losses reduce R&D budgets potentially undermining long-term innovation. Allied equipment manufacturers similarly sacrifice Chinese sales. Scientific collaboration faces restrictions slowing research progress. Global supply chain disruptions create ripple effects across chip-dependent industries. Costs are accepted as necessary for strategic objectives but remain substantial.

Strategic Implications and Future Trajectories

The October 2022 controls initiated technological decoupling with profound long-term implications. Several scenarios are plausible:

Scenario 1 (U.S. Success): Restrictions successfully constrain Chinese semiconductor capabilities at trailing-edge nodes indefinitely. American innovation continues advancing while Chinese efforts stagnate without access to critical equipment. U.S. military maintains decisive technological advantages, and American firms dominate global chip markets excluding China. Allied coordination sustains, and China accepts technological constraints.

Scenario 2 (Chinese Breakthrough): Chinese indigenous development achieves unexpected breakthroughs in manufacturing equipment (domestic EUV equivalents) or alternative architectures (RISC-V, chiplets, neuromorphic computing) that circumvent conventional constraints. Massive state investment drives rapid progress despite inefficiency. U.S. restrictions prove temporarily effective but ultimately fail as Chinese capabilities close gaps by 2030s.

Scenario 3 (Fragmented Stalemate): Restrictions prevent Chinese access to cutting-edge chips, but China develops "good enough" capabilities for most applications using trailing-edge processes. Global technology ecosystems fragment: U.S./allied advanced chips serve Western markets, Chinese chips serve domestic and Belt and Road Initiative markets. Both systems coexist with inefficiencies from duplicated R&D and lost economies of scale.

Scenario 4 (Mutual Damage): Restrictions trigger comprehensive Chinese retaliation (rare earth export prohibitions, critical mineral restrictions, pharmaceutical ingredient cutoffs). Both sides suffer substantial economic damage. Technology decoupling combines with broader economic fragmentation reducing global growth. Neither side achieves decisive advantages; both incur significant costs.

Current evidence suggests Scenario 3 (fragmented stalemate) as most likely: China develops partial capabilities sufficient for many purposes while remaining behind cutting-edge, and both ecosystems operate in parallel with strategic and economic costs. However, outcomes remain highly uncertain and dependent on Chinese indigenous development success, sustained allied cooperation, and continued American innovation investment.

2025 Policy Shift: The Trump administration's approach to semiconductor export controls illustrates the sustainability tensions inherent in Scenario 3. In January 2025, the outgoing Biden administration issued a sweeping "AI Diffusion Rule" creating a global three-tier licensing framework to prevent Chinese access to advanced chips through third countries. The Trump administration rescinded this rule, favoring a different approach: in December 2025, BIS began reviewing export license applications for Nvidia H200 and AMD MI325X chips to approved Chinese customers on a case-by-case basis — a significant relaxation from the blanket prohibitions of 2022-2024. At the same time, 140 additional PRC-linked entities were added to the Entity List, demonstrating continued pressure on specific firms even as the broader policy loosened. This oscillation between restriction and selective engagement reflects the fundamental tension between American firms' commercial interests (tens of billions in lost Chinese revenue) and strategic competition objectives — precisely the sustainability challenge this analysis identifies.

The case demonstrates how technology export controls serve strategic competition objectives while creating economic costs, enforcement challenges, and escalation risks. Success requires not just restricting adversary access but sustaining domestic innovation, maintaining allied coordination, and accepting collateral damage as acceptable price for strategic advantages. Whether these requirements can be sustained across changes in administration and over decades of competition remains the fundamental question shaping U.S.-China technology rivalry.

---

Data Sources and Further Research

Semiconductor Industry Data

Primary Sources:

• TechInsights: Semiconductor market analysis, technology assessments, and competitive intelligence on chip manufacturers

• Semiconductor Industry Association (SIA): U.S. industry statistics, policy positions, and market data

• Gartner and IC Insights: Market research, forecasts, and technology trend analysis

• Company disclosures: TSMC, Samsung, Intel, SMIC annual reports and investor presentations provide production capacity, technology node progress, and capital expenditures

Trade and Export Control Data:

• Bureau of Industry and Security (BIS): Federal Register publications of export control rules, Entity List updates, and enforcement actions

• Commerce Department: Semiconductor supply chain reviews and policy documents

• UN Comtrade: Bilateral trade flows for semiconductor equipment and chips (HS codes 8541, 8542)

Artificial Intelligence Research

Publication Databases:

• arXiv.org: Preprint server for AI/ML research papers

• Scopus and Web of Science: Citation databases for measuring research impact

• Stanford AI Index: Annual comprehensive reports on AI trends, publications, and capabilities

• Google Scholar: Author and institution publication tracking

Patent Data:

• USPTO (U.S. Patent and Trademark Office): American patent filings

• WIPO (World Intellectual Property Organization): Global patent database

• Patent analytics firms: Dimensions, PatSnap for AI-specific patent analysis

AI Chip and Computing:

• Nvidia, AMD, Intel investor relations: Product announcements and market data

• MLPerf benchmarks: Standardized AI performance metrics

• Top500 and Green500: Supercomputer rankings including AI-specific systems

Government Policy and Strategy Documents

U.S. Sources:

• National Security Strategy: High-level technology competition framing

• CHIPS and Science Act: Legislative text and Commerce Department implementation

• Inflation Reduction Act: Clean energy and battery manufacturing incentives

• National Quantum Initiative: Federal quantum technology strategy and funding

• OSTP reports: White House Office of Science and Technology Policy analyses

Chinese Sources:

• Made in China 2025: Original policy documents (English translations available from CSIS, Mercator Institute for China Studies)

• Five-Year Plans: 13th (2016-2020) and 14th (2021-2025) include technology priorities

• Ministry of Science and Technology: Policy announcements and R&D statistics (Chinese-language, partial English)

• Chinese academic journals: China Science and Technology Resources (中国科技资源导刊), Science and Technology Progress and Policy (科技进步与对策)

Think Tanks and Policy Research

U.S.-Focused:

• Center for Strategic and International Studies (CSIS): China Power Project, Strategic Technologies Program

• Center for a New American Security (CNAS): Technology and national security

• Belfer Center for Science and International Affairs: AI, cybersecurity, technology competition

• Information Technology and Innovation Foundation (ITIF): Technology policy analysis

China-Focused:

• Mercator Institute for China Studies (MERICS): European perspective on Chinese technology policy

• Rhodium Group: Chinese investment tracking and economic analysis

• Carnegie Endowment for International Peace: China technology and innovation research

• Australian Strategic Policy Institute (ASPI): Critical technology tracker

Books and Foundational Reading

Semiconductors:

• Miller, Chris. Chip War: The Fight for the World's Most Critical Technology. New York: Scribner, 2022.

• Lécuyer, Christophe and David C. Brock. Makers of the Microchip: A Documentary History of Fairchild Semiconductor. Cambridge: MIT Press, 2010.

Artificial Intelligence:

• Lee, Kai-Fu. AI Superpowers: China, Silicon Valley, and the New World Order. Boston: Houghton Mifflin Harcourt, 2018.

• Mitchell, Melanie. Artificial Intelligence: A Guide for Thinking Humans. New York: Farrar, Straus and Giroux, 2019.

Technology Competition:

• Roberts, Anthea, Henrique Choer Moraes, and Victor Ferguson. Toward a Geoeconomic Order in International Trade and Investment. Cambridge: Cambridge University Press, 2019.

• Farrell, Henry and Abraham L. Newman. "Weaponized Interdependence: How Global Economic Networks Shape State Coercion." International Security 44, no. 1 (2019): 42-79.

Chinese Technology Strategy:

• Breznitz, Dan and Michael Murphree. Run of the Red Queen: Government, Innovation, Globalization, and Economic Growth in China. New Haven: Yale University Press, 2011.

• Segal, Adam. The Hacked World Order: How Nations Fight, Trade, Maneuver, and Manipulate in the Digital Age. New York: PublicAffairs, 2016.

Databases and Tracking Tools

• China Global Investment Tracker (AEI): Chinese foreign investment by sector including technology

• CSET China AI Research Tracker: Georgetown University database of Chinese AI research

• PIIE Peterson Institute Trade Charts: U.S.-China trade flows and tariffs

• Council on Foreign Relations Global Conflict Tracker: Taiwan Strait tensions and scenarios

Recommended Monitoring for Current Developments

• Federal Register: Weekly monitoring for new export control rules

• Commerce Department newsroom: BIS announcements on Entity List and enforcement

• Industry publications: Semiconductor Engineering, EE Times, The Information (technology)

• Financial news: Financial Times, Wall Street Journal, Bloomberg for corporate developments

• Chinese media: Global Times, South China Morning Post for official perspectives (interpret carefully given state influence)

---

End of Chapter 4

Key Insights

• Semiconductors are the single most consequential chokepoint in technology competition: They are simultaneously ubiquitous (powering virtually all modern electronics), strategically critical (essential for both economic activity and military systems), and characterized by extreme production concentration. The nation controlling semiconductor supply chains holds advantages across every domain -- military, economic, and political.

• The October 2022 export controls marked a strategic pivot from "running faster" to "hobbling the competition": Previous U.S. strategy sought to maintain a relative technology lead through out-innovation. The new approach explicitly aims to slow Chinese progress through denial, accepting that commercial sales to China inevitably transfer military-relevant capabilities. This represents the most aggressive peacetime technology denial since CoCom.

• Huawei's 7nm chip achievement demonstrates that export controls slow but do not halt determined adversaries: SMIC produced a functional 7nm chip through brute-force multi-patterning using older DUV equipment, proving Chinese technical capability. However, low yields, high costs, and a technological ceiling around 5nm without EUV lithography reveal the limits of this approach, making it economically unsustainable at scale.

• AI compute has become the new strategic resource, analogous to oil in the 20th century: Training frontier AI models requires thousands of specialized chips running for months at costs exceeding $100 million. Controlling AI chip supply determines which nations can develop cutting-edge AI, creating a new form of technological inequality with direct military and economic consequences.

• Technology competition is ultimately ecosystem competition: Individual technologies can be copied or purchased, but the institutional frameworks generating sustained innovation -- research universities, venture capital, IP protection, immigration policies, entrepreneurial culture -- cannot be easily replicated. Whether China's state-directed approach can match America's market-driven ecosystem remains the fundamental open question.

• Allied coordination is essential but generates persistent tension between security imperatives and commercial interests: The Netherlands, Japan, and South Korea implemented export restrictions aligned with U.S. controls, but each country negotiated narrower restrictions than Washington preferred, reflecting the difficulty of sustaining cooperation when allies bear disproportionate commercial losses.

• Collateral damage from technology restrictions undermines the innovation base they aim to protect: American semiconductor and AI chip firms lose tens of billions in annual Chinese revenue, reducing R&D budgets that fund future innovation. The paradox is that technology denial may weaken the coercer's long-term competitive position even as it constrains the adversary's near-term capabilities.

Discussion Questions

1. National Security Advisor Jake Sullivan articulated a shift from maintaining a "relative" technology advantage to establishing "as large a lead as possible." Is this objective achievable and sustainable, or does it set the United States on an unsustainable path of escalating restrictions that eventually fragment the global innovation ecosystem?

2. Nvidia initially designed "China-compliant" AI chips (A800, H800) that marginally reduced specifications to fall below export control thresholds while maintaining substantial capabilities. This cat-and-mouse dynamic between industry and regulators is a recurring feature of export controls. How should policymakers balance the need for clear, enforceable rules with the inevitability of specification gaming by commercially motivated firms?

3. China's state-directed semiconductor development has consumed over $100 billion with mixed results. Under what conditions might massive state investment in technology eventually succeed despite market-oriented economies' historical advantages in innovation? What features of the semiconductor industry make it more or less amenable to state-directed development?

4. The chapter presents AI as a dual-use technology where the same chips training consumer chatbots also train military targeting systems. Given this inseparability, is it possible to design export controls that meaningfully restrict military AI development without crippling legitimate commercial and scientific applications? What would such controls look like?

5. If semiconductor export controls successfully constrain Chinese AI and computing capabilities for the next decade, but China eventually achieves self-sufficiency in semiconductor equipment, will the United States be better or worse off than if it had maintained commercial engagement? How should policymakers weigh short-term denial benefits against long-term relationship costs?

---

Tabletop Exercise: The tabletop exercise for this chapter — AI Export Control Dilemma — can be found in Appendix A: Tabletop Exercises.

---

References and Further Reading

Allen, Gregory C. "Choking Off China's Access to the Future of AI." Center for Strategic and International Studies, October 2022.

Allen, Gregory C., and Emily S. Weinstein. "Huawei's Mate 60 Pro: China's Semiconductor Breakthrough." CSIS Commentary, September 2023.

Khan, Saif M., Alexander Mann, and Dahlia Peterson. "The Semiconductor Supply Chain: Assessing National Competitiveness." Center for Security and Emerging Technology, January 2021.

Miller, Chris. Chip War: The Fight for the World's Most Critical Technology. Scribner, 2022.

Rasser, Martijn, et al. "The CHIPS and Science Act: Here Comes the Hardest Part." Center for Strategic and International Studies, August 2022.

SIA (Semiconductor Industry Association). "2023 State of the U.S. Semiconductor Industry." June 2023.

Sullivan, Jake. "Remarks by National Security Advisor Jake Sullivan at the Special Competitive Studies Project Global Emerging Technologies Summit." White House, September 16, 2022.

Triolo, Paul, and Robert Greene. "Will China's Chip Breakthrough Undermine U.S. Tech Restrictions?" Carnegie Endowment for International Peace, September 2023.

U.S. Department of Commerce, Bureau of Industry and Security. "2022 Export Controls: Taking the Long View." October 2022.