The Silicon Sovereignty Race: AI Hardware Supply Chains and the New Technology Frontier

The semiconductor became a geopolitical instrument before most policymakers fully understood what a semiconductor was. The recognition arrived abruptly, compressed into the late 2010s by a collision of forces: Chinese state investment in chip manufacturing, the dominant position of Taiwanese foundries in advanced logic, American export control regimes that weaponized the global chip supply chain, and the emergent understanding that artificial intelligence — the technology reshaping economic and military power — was ultimately a hardware competition. Silicon had become sovereign territory.

This analysis examines the AI hardware sovereignty race in its full complexity: the technical realities that constrain national ambitions, the geopolitical dynamics that motivate them, the industrial policies that express them, and the strategic implications for the nations, companies, and institutions navigating a world in which the ability to design and manufacture advanced semiconductors is increasingly treated as equivalent to the ability to exercise independent national power. The argument is that the race for AI hardware sovereignty is not a temporary disruption but a permanent feature of the competitive landscape, and that its implications extend far beyond the technology industry to reshape the architecture of global power.

Why Hardware Matters for AI

Understanding the geopolitics of AI hardware requires understanding the technical dependencies that make hardware so consequential. Artificial intelligence, at its current state of development, is extraordinarily computationally intensive. Training a large language model at frontier scale requires tens of thousands of specialized processors running continuously for weeks or months. Inference — deploying a trained model to serve users — requires continuous compute infrastructure at enormous scale. The economics of AI are, in substantial part, the economics of compute.

The specialized processors required for AI workloads — primarily graphics processing units (GPUs) designed for parallel computation and more recently custom application-specific integrated circuits (ASICs) — have distinctive technical requirements. They require extraordinarily dense transistor geometries, currently at the 3-nanometer to 5-nanometer scale, achievable only through the most advanced lithography processes. They require high-bandwidth memory architectures that push the limits of packaging technology. They require advanced interconnect technologies for multi-chip configurations. And they require the accumulated manufacturing expertise — the process knowledge built through years of production experience — that cannot be purchased or transferred quickly.

The gap between the ability to design an advanced AI chip and the ability to manufacture it at scale is enormous. Nations that can design chips but cannot manufacture them have theoretical capability without operational independence. The strategic significance is in the full stack.

This technical reality creates a profound dependency structure. The global semiconductor supply chain is among the most complex industrial systems ever assembled, spanning materials, equipment, design tools, intellectual property, and manufacturing across dozens of countries. No single country controls the full stack. But some countries control components of the stack that are so strategically positioned — so difficult to replace and so essential to the whole — that their control constitutes effective leverage over the entire system.

The Chokepoint Architecture

The global semiconductor supply chain has several structural chokepoints that define where geopolitical leverage actually resides.

Extreme ultraviolet lithography (EUV) equipment is the technology that enables manufacturing at the most advanced node sizes. Essentially all of this equipment is produced by a single Dutch company, ASML, whose machines are the result of decades of collaborative development involving components from hundreds of suppliers across Europe, Japan, and the United States. ASML's near-monopoly on EUV equipment is not an accident or an anticompetitive outcome — it reflects the extraordinary technical difficulty of the problem. No alternative EUV supplier exists or is close to existing.

Electronic design automation (EDA) tools, the software used to design chips, is dominated by three American companies: Synopsys, Cadence Design Systems, and Siemens EDA. Advanced chip design is essentially impossible without access to these tools. Their position represents a control point of immense strategic significance that is relatively invisible compared to the more dramatic hardware competition.

Advanced foundry manufacturing is dominated by Taiwan Semiconductor Manufacturing Company (TSMC), which manufactures semiconductors for the vast majority of the world's chip designers, including NVIDIA, AMD, Apple, Qualcomm, and many others. Samsung Semiconductor holds a secondary position in advanced foundry manufacturing. No other company operates at the leading edge at meaningful scale. Intel's foundry ambitions, while backed by substantial public and private investment, remain works in progress.

Specialty materials and process chemicals represent distributed but significant chokepoints. Japan dominates the production of many photoresists, etchants, and other specialty materials essential to semiconductor manufacturing. Disruption of Japanese supply would cascade through the entire global manufacturing system.

The American Export Control Revolution

The Biden administration's October 2022 export control measures, expanded significantly in subsequent rounds through 2023 and 2024, represented a fundamental shift in American industrial and technology policy. The measures restricted the export of advanced semiconductors, chip manufacturing equipment, and related software to China, and — in a particularly consequential provision — restricted the ability of American citizens and permanent residents to support Chinese chip manufacturing at leading-edge nodes.

The strategic logic was explicit and unprecedented. Rather than merely preventing the export of specific military-use technologies, the measures aimed to degrade China's entire ability to develop and produce advanced semiconductors over time. The stated objective was to prevent China from obtaining or manufacturing the chips needed to train and deploy frontier AI systems at scale. This was not export control as it had been practiced — preventing the diversion of specific dual-use items to specific prohibited end users. This was export control as industrial strategy: deliberately managing global technological development to prevent a strategic competitor from closing a capability gap.

The measures were extraordinary in several respects. They applied extraterritorially through the foreign direct product rule, which subjects foreign-made products incorporating American technology to American jurisdiction. They effectively compelled allied nations and their companies to participate in the control regime or lose access to American technology and markets. And they were explicitly justified not by specific military threats but by the general principle that denying China access to advanced compute would sustain American advantage in AI capabilities with compounding military and economic implications.

The response from China was accelerated investment in domestic alternatives, a public commitment to semiconductor self-sufficiency, and an increasingly sophisticated effort to circumvent controls through third-country diversion, shell company procurement, and indigenous development programs. None of these responses has yet closed the gap at the leading edge. All of them are making the gap harder to maintain.

The American export control regime is, in effect, a bet that the technological gap in semiconductor manufacturing is large enough, and the difficulty of closing it great enough, that aggressive controls can preserve American advantage in AI hardware for long enough to matter strategically. Whether that bet is correct will be determined in the next five to seven years.

China's Semiconductor Strategy: Ambition, Investment, and Limitation

China's pursuit of semiconductor self-sufficiency predates the American export controls but has been dramatically accelerated by them. The Made in China 2025 program, launched in 2015, identified semiconductors as a priority sector. The National Integrated Circuit Industry Investment Fund (the "Big Fund"), which has disbursed hundreds of billions of renminbi over two investment rounds, has financed a broad expansion of Chinese semiconductor capacity, design capability, and manufacturing equipment development.

The limitations of Chinese semiconductor ambitions are real and significant, but they have been systematically underestimated in American strategic assessments that have relied too heavily on the assumption that the leading-edge gap is insurmountable.

SMIC's progression. Semiconductor Manufacturing International Corporation, China's leading foundry, demonstrated in 2023 that it had achieved production at 7-nanometer-equivalent node sizes, believed to be using a technique called deep-UV multi-patterning rather than EUV. The yield rates and cost structure of SMIC's production at this node are believed to be well below TSMC's, but the technical achievement was significant and demonstrated that the gap between China's manufacturing capability and the global frontier, while real, is not as large as it was commonly assumed.

Huawei's Kirin 9000S. The Mate 60 Pro smartphone, introduced in late 2023, contained a Kirin 9000S chip manufactured by SMIC. Its performance demonstrated that Chinese AI chip capability, while behind NVIDIA's leading products, is advancing at a pace that challenges the assumption that export controls will maintain a permanent and decisive gap.

Equipment development. Chinese investment in domestic semiconductor equipment is substantial and growing. Companies including NAURA, AMEC, and Shanghai Micro Electronics Equipment Group (SMEE) are progressing toward domestic alternatives for deposition, etching, and — most ambitiously — lithography equipment, though they remain well behind ASML in EUV capability.

The trajectory matters more than the current position. China is moving up the capability curve in semiconductor manufacturing, albeit from a substantial lag position. The speed of that progression, relative to the speed of the American frontier's advance, will determine whether the export control strategy succeeds in its objectives.

The Chinese AI Chip Design Ecosystem

Separate from the manufacturing constraint is the question of China's AI chip design capability. This is, in important respects, more advanced and less constrained than manufacturing. Huawei's Ascend series of AI accelerators, Cambricon's MLU series, and a range of smaller AI chip design companies have produced increasingly capable processors. Design is less dependent on the manufacturing chokepoints — though it is ultimately limited by manufacturing constraints in that designs can only be realized in silicon at whatever node the available foundries can produce.

The combination of advancing domestic manufacturing capability and a vibrant domestic AI chip design ecosystem means that the American strategy of maintaining AI hardware superiority through supply chain controls faces a race against time. The controls may successfully prevent China from obtaining the most advanced American chips. They are less certain to prevent China from developing domestic alternatives that are sufficient for many AI applications, even if not at full parity with the leading edge.

Taiwan: The Island at the Center of Everything

No analysis of AI hardware sovereignty can avoid Taiwan's singular strategic position. TSMC's dominance of advanced semiconductor foundry manufacturing means that Taiwan hosts the most strategically consequential manufacturing capability on earth. Advanced chips for American defense systems, commercial AI infrastructure, consumer electronics, and critical infrastructure globally depend on manufacturing processes that can only be performed at TSMC fabs on the island of Taiwan, an island whose political status is contested by the People's Republic of China.

The strategic implications are staggering in their scope. A disruption to TSMC's operations — whether through military action, political crisis, natural disaster, or industrial sabotage — would not merely inconvenience the electronics industry. It would halt production of advanced semiconductors globally, with cascading effects across every sector that depends on them, which is to say, every sector.

TSMC has invested substantially in geographic diversification under pressure from American, Japanese, and European governments. A facility in Arizona is under construction, with substantial American government subsidization under the CHIPS Act. Facilities in Japan, supported by Japanese government investment, are progressing. A potential European facility has been discussed. These investments are expanding TSMC's geographic footprint in ways that reduce, at the margin, the concentration of critical manufacturing in Taiwan.

But the scale of these investments relative to TSMC's Taiwan capacity is important to keep in mind. The Arizona facility, when complete, will represent a small fraction of TSMC's total capacity. The most advanced process nodes, and the full depth of manufacturing expertise embodied in TSMC's workforce and process knowledge, will remain concentrated in Taiwan for the foreseeable future. Geographic diversification is a risk reduction measure. It does not eliminate the Taiwan concentration risk.

The Taiwan question is not merely a political and military issue about the fate of a democratic island. It is a question about the most critical single concentration point in the global technology supply chain. Any analysis of Taiwan's strategic status that ignores the semiconductor dimension is incomplete. Any analysis of AI hardware sovereignty that ignores Taiwan is incomplete.

The CHIPS Act and Allied Industrial Policy

The American CHIPS and Science Act, passed in 2022 with bipartisan support, committed approximately fifty-two billion dollars in direct subsidies for domestic semiconductor manufacturing and research, supplemented by investment tax credits that roughly doubled the effective value of the support. The CHIPS Act was explicitly motivated by national security concerns about semiconductor geographic concentration and Chinese competitive advances.

Similar programs have been launched or expanded in the European Union (the European Chips Act), Japan (substantial foundry subsidies to attract TSMC and other manufacturers), South Korea, India, and a range of other countries. The global policy response to semiconductor concentration risk has been rapid and substantial by the standards of industrial policy.

The effectiveness of these programs in actually reshuffling the competitive landscape will take years to assess. Building semiconductor manufacturing capability is a decade-long undertaking. The talent requirements, supply chain requirements, and process development requirements for leading-edge foundry manufacturing are not satisfied by capital investment alone. Whether government subsidies can accelerate the development of genuine manufacturing capability, or merely construct facilities that depend on imported expertise and equipment indefinitely, is an open question.

The AI Compute Race: Implications for National Strategy

The link between AI hardware and national strategic power is not merely theoretical. AI systems trained on sufficient compute at sufficient scale demonstrate emergent capabilities that are strategically relevant across economic, military, and intelligence dimensions. The ability to field frontier AI systems depends, at present, on the ability to access massive quantities of advanced compute. Nations and actors that cannot access frontier compute cannot develop or deploy frontier AI systems.

This dependency creates a new dimension of strategic competition that is structurally similar to, but analytically distinct from, prior technology competition. Unlike nuclear weapons, AI systems are commercial technologies with civilian applications that create strong economic incentives for development independent of military applications. Unlike software, AI capability is hardware-constrained in ways that create real barriers to entry based on physical capital. Unlike traditional industrial competition, the competitive dynamics of AI development involve network effects, data advantages, and feedback loops that can create rapid and durable capability leads.

Military Implications of AI Hardware Gaps

Military applications of AI are advancing rapidly and will be determinative in future conflicts in ways that are not yet fully reflected in strategic doctrine or force structure. Autonomous systems, intelligence analysis, electronic warfare, logistics optimization, and command and control all have AI dimensions whose effectiveness will be constrained by the quality and quantity of available AI hardware.

A nation that cannot manufacture or obtain advanced AI chips is a nation whose military AI development is constrained by supply chain dependencies that an adversary can potentially exploit. This is not hypothetical. American export controls on advanced semiconductors to China are explicitly designed to create and maintain this kind of constraint. The military logic of hardware sovereignty is that true strategic independence in AI-relevant military capability requires domestic production capability for the hardware that AI runs on.

A military AI capability is not sovereign if the hardware it depends on is manufactured by a potential adversary or a third party subject to adversary pressure. Sovereignty in AI capability requires sovereignty in at least the critical elements of AI hardware.

Economic Implications: Who Owns the Infrastructure of Intelligence

Beyond military applications, advanced AI will increasingly constitute the infrastructure through which economic activity is organized, optimized, and conducted. Companies and nations that control the infrastructure of AI — the compute, the models, the platforms — will have structural advantages in productivity, innovation, and competitive positioning that compound over time.

This creates economic sovereignty concerns that parallel military ones. Nations that host their AI infrastructure on foreign-controlled hardware and cloud services, trained on foreign-developed foundation models, are in a position of dependency that may have significant economic implications. The ability to operate AI systems independently — with hardware that can be procured domestically or from trusted allies, models that can be trained and fine-tuned domestically, and infrastructure that is not subject to foreign control — is increasingly treated as an element of economic sovereignty analogous to energy sovereignty.

The recent surge in "sovereign AI" initiatives — programs in France, Germany, Japan, the UAE, Saudi Arabia, and numerous other countries to build or procure domestically controlled AI infrastructure — reflects this recognition. The strategic logic is clear even when the operational execution is imperfect.

NVIDIA's Structural Position and Its Implications

No analysis of AI hardware geopolitics is complete without examining NVIDIA's extraordinary structural position. Through a combination of early investment in GPU architecture, the CUDA software ecosystem built over more than a decade, and the fortunate alignment of GPU architecture with the specific computational requirements of deep learning, NVIDIA has achieved a position in AI computing hardware that is, in important respects, more dominant than any single technology company has achieved in any strategic technology in recent memory.

NVIDIA's market share in the AI accelerator market has, at various points, exceeded ninety percent for frontier training workloads. The CUDA software ecosystem — the programming model, libraries, tools, and accumulated developer knowledge built around NVIDIA hardware — represents a switching cost that has proven durable even as alternative hardware has become available. Organizations that have invested in CUDA-based AI development pipelines face substantial costs in migrating to alternative hardware, even when the alternative hardware offers competitive performance at lower cost.

This position creates strategic dependencies that have both commercial and geopolitical dimensions. Nations and organizations that depend on NVIDIA hardware for AI infrastructure are dependent on a single American company whose products are subject to American export control jurisdiction. The export controls that have restricted NVIDIA's sales to China have demonstrated this dependency concretely: the restrictions required Chinese AI developers to either use domestically developed alternatives with lower performance or obtain chips through indirect channels.

The durability of NVIDIA's position is the subject of genuine strategic debate. The emergence of capable alternative AI accelerators — from Google (TPUs), Amazon (Trainium/Inferentia), Microsoft (Maia), Meta (MTIA), and a range of start-ups — suggests that the NVIDIA position may be less permanent than it appears. But the CUDA ecosystem's switching costs are real, NVIDIA's architectural pace of development has been relentless, and the company's scale advantages in manufacturing access and ecosystem investment are substantial.

NVIDIA's position in AI hardware is analogous to a critical infrastructure monopoly in a strategic sector. This creates both commercial risks for customers and geopolitical risks for nations whose AI ambitions are mediated through a single supplier subject to a single government's jurisdiction.

The Sovereign AI Hardware Stack: What It Requires

Nations serious about AI hardware sovereignty — the ability to develop and deploy frontier AI capabilities without critical dependencies on adversary or unreliable third-party supply chains — face a daunting technical and industrial challenge. The full AI hardware sovereignty stack spans multiple layers, each requiring distinct capabilities.

Foundation model training infrastructure requires access to large clusters of advanced AI accelerators, high-bandwidth networking (InfiniBand or equivalent), and the operational expertise to manage large-scale distributed training runs. This layer is relatively achievable for well-resourced nations through procurement, even if domestic manufacturing is not available, provided export control access is maintained.

Advanced chip manufacturing is the hardest layer. Achieving domestic production at the process nodes required for competitive AI hardware — currently 3nm to 7nm for leading products — requires a full ecosystem of manufacturing equipment, materials, process chemistry, and skilled workforce that takes a decade or more to build. No nation other than Taiwan, the United States (with TSMC's Arizona facility), South Korea, Japan, and potentially the Netherlands has a realistic path to near-term advanced node manufacturing at scale.

Chip design capability is more achievable than manufacturing. High-quality chip design, including the architecture of AI accelerators, is primarily a human capital and software tool problem. Nations can develop or import the talent and tool access needed for competitive chip design on a shorter timeline than manufacturing. The limitation is that design capability without manufacturing capability is dependent on foundry relationships that may be constrained by export controls or geopolitical circumstances.

EDA tools and IP represent a hidden dependency layer. The design tools and intellectual property blocks that underpin modern chip design are substantially American-controlled. This creates a vulnerability that is less visible than hardware manufacturing but potentially as constraining.

Materials and equipment supply chains for semiconductor manufacturing are globally distributed but concentrated in specific chokepoints in the Netherlands, Japan, and the United States. Achieving full supply chain independence requires either developing domestic alternatives to these chokepoints — an enormous industrial undertaking — or ensuring reliable access through treaties, investments, and relationship management with allied nations.

Europe's Semiconductor Dilemma

Europe's position in the AI hardware competition is distinctive and in some respects paradoxical. Europe hosts ASML, which makes the most critical single piece of manufacturing equipment in the global semiconductor supply chain. Europe hosts Infineon, STMicroelectronics, and NXP Semiconductors, which are world-leading in automotive, industrial, and power semiconductors. Europe has strong materials and equipment companies. And yet Europe has essentially no presence in leading-edge logic manufacturing — the production of the advanced chips needed for AI — and limited presence in AI chip design.

The European Chips Act represents an ambitious effort to address this gap. The program aims to increase Europe's share of global semiconductor production to twenty percent by 2030, from a current position of roughly ten percent. The twenty percent target has been widely criticized as unrealistic given the investment scale required and the lead times involved. But the strategic direction is correct: Europe recognizes that dependency on Taiwanese, Korean, and American manufacturers for advanced logic represents a strategic vulnerability.

The deeper challenge for Europe is that industrial policy in a heterogeneous multi-member union is harder to execute than in a unitary state. The allocation of subsidies, the coordination of national programs, and the management of competition concerns among member states all complicate the policy execution challenge. European chip policy is further complicated by the fact that ASML's export control position — whether and how ASML equipment can be sold to China — is a point of geopolitical friction between Europe and the United States, with significant commercial interests at stake.

Europe's semiconductor paradox is that it hosts the company that makes the most strategically important piece of equipment in global chip manufacturing, and yet has minimal presence in the advanced chip manufacturing that equipment enables. This disconnect reflects decades of underinvestment in manufacturing capability that the European Chips Act is trying to reverse on an accelerated timeline.

The Longer Game: Beyond the Current Transition

The AI hardware sovereignty race, as currently constituted, is a competition organized around a specific technological paradigm: large-scale GPU-based training of neural networks with dense transformer architectures. This paradigm has dominated AI development since approximately 2017. It is not permanent.

Several technological developments could alter the competitive dynamics significantly.

Neuromorphic and analog computing approaches that more closely mimic biological neural architectures may offer dramatically superior energy efficiency for specific AI workloads. Several major research programs are investing heavily in these approaches. If they mature into competitive platforms for AI inference or eventually training, the manufacturing requirements and competitive dynamics could shift significantly.

Photonic computing uses light rather than electrons for computation and offers potential efficiency advantages for specific linear algebra operations central to AI. Photonic computing companies are advancing, though current systems remain special-purpose and limited in scope.

Memory-centric computing architectures that reduce the data movement overhead of conventional von Neumann architectures may offer substantial efficiency improvements for AI workloads. This could alter the balance of value and competitive advantage between processor manufacturers and memory manufacturers.

Quantum computing, while unlikely to be directly relevant to training neural networks in the near term, could become relevant for specific AI-adjacent computational tasks — optimization problems, simulation — on a longer timeline.

These developments do not undermine the current strategic urgency. The AI hardware competition of the next five to ten years will be determined by the current technological paradigm. Nations and companies that establish strong positions in that paradigm will have advantages that compound even through technological transitions. But the longer-term landscape is genuinely uncertain, and strategies that optimize excessively for current architectures may prove brittle.

Strategic Implications for Decision-Makers

Several high-confidence conclusions emerge from this analysis for decision-makers navigating the AI hardware sovereignty landscape.

The hardware dependency is structural and enduring. As long as AI capability is compute-constrained, which will be true for at least the next decade, hardware sovereignty is genuine strategic sovereignty. Decision-makers who treat AI as a primarily software problem are missing the dependency that constrains everything else.

The export control regime will escalate. American export controls on advanced semiconductors are a policy choice, not a permanent equilibrium. They create pressure for Chinese self-sufficiency investment, they impose costs on allied nations with China exposure, and they create commercial incentives for workarounds. The regime will evolve — potentially toward tighter controls, potentially toward negotiated alternatives, almost certainly toward increased enforcement complexity.

Allied coordination is essential and difficult. The chokepoints in the AI hardware supply chain span multiple allied nations. Effective hardware sovereignty strategy for any single nation requires coordination with allies on export controls, industrial policy, and supply chain resilience. This coordination is more difficult than unilateral action and more effective.

Diversification reduces but does not eliminate concentration risk. The investments in TSMC Arizona, TSMC Japan, and the range of other geographic diversification efforts will reduce the catastrophic concentration risk centered on Taiwan. They will not eliminate it, and they will not produce genuinely independent manufacturing capability outside of Taiwan on a near-term timeline.

The race is real, the outcome is uncertain. China's semiconductor ambitions face real constraints from export controls, technical gaps, and manufacturing complexity. They have also demonstrated more progress than optimistic assessments of the American control strategy anticipated. The outcome of the AI hardware sovereignty race is not predetermined by current advantage.

Nations that wait for market forces to solve their hardware sovereignty challenges will find, as they have in energy, that dependency is most acute precisely when the market has least incentive to solve it. The time for strategic investment is before the dependency creates leverage, not after.

Conclusion: Silicon as Strategic Territory

The AI hardware sovereignty race is, at its core, a competition about which nations will have genuine freedom of action in the AI era. That freedom depends not merely on having access to AI systems but on having independent control of the infrastructure those systems require — the chips, the manufacturing, the design tools, the materials, and the talent.

No nation currently has full stack independence. The United States comes closest, but depends on Taiwanese manufacturing for leading-edge production and on complex global supply chains for materials and equipment. Taiwan has unparalleled manufacturing capability but depends on American design tools and market access. China has scale, investment, and growing technical capability but faces deliberate supply chain constraints. Europe has critical chokepoint technology but limited manufacturing presence at the leading edge.

The strategic responses to this situation — CHIPS Acts, export controls, sovereign AI initiatives, allied coordination frameworks — reflect a recognition that the dependencies embedded in the current AI hardware supply chain are not commercially acceptable from a national security perspective. That recognition has arrived late and will be tested against the pace of Chinese indigenous development, the resilience of the Taiwan concentration point, and the speed at which industrial policy can actually build manufacturing capability that competes with decades of accumulated TSMC expertise.

The companies and governments that navigate this transition most effectively will be those that understand both the technical realities and the geopolitical dynamics — that recognize where genuine chokepoints exist, where leverage actually resides, and where the gap between ambition and capability is likely to persist despite political will and financial investment. Silicon, in the most literal sense, has become strategic territory, and its governance will shape the distribution of power in the coming decade as surely as oil shaped the previous one.

Sources & References

Semiconductor Industry Association (SIA) Annual Reports ASML Annual Reports and Investor Presentations TSMC Annual Reports and Investor Presentations NVIDIA Annual Reports and Investor Presentations U.S. Bureau of Industry and Security Export Control Rules and Guidance Congressional Research Service — Semiconductors and the CHIPS Act Center for Strategic and International Studies — Technology and National Security Program Brookings Institution — Technology Policy Reports MIT Technology Review IEEE Spectrum The Economist Financial Times Wall Street Journal Nikkei Asia South China Morning Post War on the Rocks — Technology and Defense Analysis Georgetown University Center for Security and Emerging Technology (CSET) Hudson Institute — Technology and Geopolitics Research Rhodium Group — China Technology Research Paul Triolo, Eurasia Group — Technology Policy Research SemiAnalysis (industry analysis publication) Bits and Chips (European semiconductor industry publication) U.S. Department of Commerce CHIPS Program Office European Commission — European Chips Act Background Japan Ministry of Economy, Trade and Industry — Semiconductor Strategy Korea Institute for Industrial Economics and Trade — Semiconductor Reports