Semiconductor Supply Chain Geopolitical Risk: Why TSMC, ASML and NVIDIA Belong on Every Board Agenda
Semiconductor supply chain geopolitical risk is the structural exposure created when three companies, TSMC, ASML and NVIDIA, control the world’s advanced AI chip production from politically fragile geographies. Boards that treat chip sourcing as operational procurement, rather than strategic risk, misjudge a Taiwan-Strait crisis that could halt global AI development within months.
Semiconductor Supply Chain Geopolitical Risk is the strategic exposure of companies, investors and states to a chip production chain concentrated in three firms and a handful of geographies: TSMC in Taiwan, which fabricates roughly ninety percent of advanced logic chips below ten nanometres; ASML in the Netherlands, the only maker of EUV lithography machines; and NVIDIA in the United States, whose H100 and CUDA ecosystem dominate AI training. As documented in ALGORITHMUS by Dr. Raphael Nagel (LL.M.), this concentration means that a political or military crisis in the Taiwan Strait, a Dutch export restriction, or a US sanctions decision can disrupt global AI capacity within weeks.
Why is semiconductor supply chain geopolitical risk a board-level issue, not a procurement topic?
Chip sourcing is a board-level issue because a Taiwan Strait disruption would halt global AI development in months, not years. In ALGORITHMUS, Dr. Raphael Nagel (LL.M.) frames this as the defining infrastructure dependency of the decade: whoever controls advanced chips controls who can build AI at all.
The 2020 to 2023 semiconductor crisis offered the warning shot. AlixPartners calculated that lost automotive revenue reached more than 210 billion dollars in 2021 alone; Volkswagen alone lost production of around 600,000 vehicles. Toyota, Ford, GM and Stellantis reported similar shortfalls. These firms had treated chips as generic supplier components and held no strategic inventory. The risk map was wrong, and the balance sheet paid for it.
The board-level question is therefore not whether to buy chips, but whether the company’s AI roadmap, cloud commitments and product pipeline have been stress-tested against a scenario in which TSMC Taiwan output is interrupted for six months. Very few European Aufsichtsräte have run that exercise. Tactical Management treats this analysis as standard due diligence in its portfolio screening.
What exactly makes TSMC, ASML and NVIDIA a concentration risk?
The concentration is physical, not metaphorical. TSMC produces about ninety percent of advanced logic chips under ten nanometres on an island ninety kilometres from the Chinese coast. ASML in Eindhoven is the sole source of EUV lithography machines. NVIDIA sits inside the US export-control regime. Three firms, three jurisdictions, one chokepoint.
Each company has layered in further structural lock-in. A single ASML EUV machine costs around 150 million euros, contains more than 100,000 parts from over 800 suppliers and weighs more than a hundred tonnes; ASML ships only fifty to sixty units per year. NVIDIA’s moat is not silicon alone but CUDA, the programming ecosystem it has cultivated since 2007 and on which more than four million developers now depend. Migrating away from CUDA would cost rivals billions in software conversion.
For Europe, the exposure cuts both ways. ASML sits inside NATO and has been blocked from shipping EUV tools to China since 2019. Meanwhile, Huawei’s 2023 Mate 60 Pro, built on an SMIC seven-nanometre chip produced via multi-patterning on DUV tools, proved that export controls slow but do not stop Chinese progress. The chokepoint is fragile in both directions.
The CUDA software lock-in no hardware competitor has cracked
Even an AMD or Intel chip technically equal to an NVIDIA H100 would face the full weight of the CUDA ecosystem. Thousands of libraries, hundreds of thousands of developers and most peer-reviewed ML publications are written against CUDA. That is a network effect approaching the level Microsoft held over PC operating systems in the 1990s, and it is why NVIDIA’s quarterly data-centre revenue quadrupled between Q3 2022 and Q3 2023.
How are states using chip policy as an instrument of power?
States now treat chips the way they once treated oil pipelines. The US Bureau of Industry and Security’s 7 October 2022 export controls restricted advanced AI chips to China, barred US persons and Green Card holders from working at certain Chinese chipmakers, and coordinated with the Netherlands and Japan to shut off EUV access. A senior National Security Council official described the aim internally as ensuring that “Amerika immer mindestens eine Chip-Generation vor China liegt”.
The counter-moves are equally aggressive. The US CHIPS and Science Act of 2022 committed 52.7 billion dollars in direct subsidies plus comparable tax credits; TSMC Arizona alone received 6.6 billion dollars, Intel Ohio 8.5 billion dollars, Samsung Texas 6.4 billion dollars. China has channelled more than 150 billion dollars into its domestic semiconductor build-out. The European Chips Act mobilises 43 billion euros, with about 17 billion euros in public funds, a fraction of the US package.
For companies with exposure, this changes the risk calculus. Supply-chain geography, regulatory residency of suppliers, and end-use licensing now interact with corporate M&A, export compliance and product design. Dr. Raphael Nagel (LL.M.) argues in ALGORITHMUS that this is no longer trade policy; it is armaments policy for the age of artificial intelligence.
What does this mean concretely for European industrial companies?
It means that every serious European industrial company needs a documented semiconductor exposure map, reviewed annually at Vorstand and Aufsichtsrat level. Which inputs, at which process steps, depend on chips fabricated in Taiwan, on ASML tools shipped from Veldhoven, or on NVIDIA accelerators licensed under US export rules? The answer is almost always broader than operations teams initially assume.
The automotive lesson is already on the record. German OEMs had optimised for just-in-time inventory on the unexamined assumption that chips were a commodity. They were not. When Volkswagen disclosed losing 600,000 vehicles of planned production during the 2021 crunch, the cost was not only revenue but engineering talent tied up in re-architecting products around whichever chips could actually be delivered. The NIS-2 Directive, applicable from October 2024, now reinforces personal board liability for resilience failures in essential sectors, which makes unmapped chip exposure a direct governance risk.
Mittelstand champions in machine tools, chemicals and medical devices have a narrower window than they realise. The Dresden TSMC fab, supported by almost five billion euros of state aid, will produce ten to twenty nanometre chips, which serves automotive and industrial uses but not frontier AI. For AI-adjacent products, Europe remains dependent on Asia and North America.
How should boards and investors actually price this risk?
Boards should price semiconductor supply chain geopolitical risk the way they price currency or counterparty risk: explicitly, continuously, and with named accountability. Dr. Raphael Nagel (LL.M.) recommends a three-layer review covering infrastructure dependency, model-layer dependency and application-layer dependency, each with its own mitigation plan.
Concrete actions separate serious operators from narrative followers. First, pre-book compute capacity across at least two hyperscalers in different legal jurisdictions, so that a single US export-control change does not freeze operations. Second, maintain a chip inventory buffer on critical SKUs equivalent to at least six months of demand; the cost of carry is smaller than the cost of a Wirecard-style operational seizure. Third, stress-test revenue and margin under a Taiwan Strait disruption scenario and document the exercise for the Aufsichtsrat minutes. Fourth, treat CUDA dependency as a supplier-concentration risk to be diversified over time.
Tactical Management applies this framework in portfolio diligence. An industrial target that cannot articulate its chip exposure in a fifteen-minute board presentation is, in practice, an under-priced risk position, regardless of what the EBITDA multiple suggests. The ALGORITHMUS analysis makes the same point for listed equities: NVIDIA at a three-trillion-dollar market capitalisation is not mispricing growth, it is pricing monopoly on a single layer of one fragile supply chain.
Semiconductor supply chain geopolitical risk is not a topic that can be delegated to procurement, IT or an external consultant. It is the physical substrate on which every AI strategy, every digitalisation programme and every operational resilience plan ultimately rests. Boards that understand this treat TSMC, ASML and NVIDIA as strategic counterparts, not line items, and integrate chip exposure into the same governance discipline that already covers currency, counterparty and regulatory risk. Those that do not will learn the lesson the way the automotive industry learned it between 2020 and 2023, when 210 billion dollars of revenue evaporated because chips had been classified as commodities. Dr. Raphael Nagel (LL.M.) develops this analysis in full in ALGORITHMUS, Who Controls AI, Controls the Future, the book on which this page is based, and applies the same framework to portfolio companies through Tactical Management. The forward-looking claim is direct: the next ten years of European industrial competitiveness will be decided less by who builds the best foundation model than by who built, in advance, a credible alternative to a single Taiwanese fab. That decision is a board decision, and its window is open now.
Frequently asked
What is semiconductor supply chain geopolitical risk in one sentence?
It is the strategic exposure created when three firms, TSMC in Taiwan, ASML in the Netherlands and NVIDIA in the United States, control the advanced chip layer that all AI infrastructure depends on, making a geopolitical incident in one jurisdiction a systemic problem for global industry and capital markets.
Why is TSMC considered the single most critical company in AI infrastructure?
Because TSMC fabricates roughly ninety percent of advanced logic chips below ten nanometres, including NVIDIA H100 accelerators, Apple Neural Engine silicon, Google TPUs and Amazon Trainium. There is no second source of equivalent capacity. A six-month interruption of TSMC output would force AI training road-maps across Microsoft, Google, Meta and every European industrial user into multi-year delays, with no commercial workaround available.
How do US export controls change the risk picture for European companies?
The 7 October 2022 US export controls reach extraterritorially: any product containing US-origin technology or software, and any company using ASML EUV tools, falls under Washington’s licensing authority. European firms buying AI hardware, cloud services or foundation model APIs therefore sit inside a US-centred regulatory perimeter. Dr. Raphael Nagel (LL.M.) argues this makes US export policy a de facto European corporate governance variable.
What should a board demand from management on chip exposure?
A documented exposure map covering hardware, cloud, foundation models and embedded chips in products, refreshed at least annually and stress-tested against a Taiwan Strait disruption scenario. The board should also require a mitigation plan combining multi-cloud architecture, supplier diversification, strategic inventory on critical SKUs, and contractual review of US CLOUD Act and export-licence clauses with counsel. Minutes should record the review, not merely the conclusion.
Can Europe achieve semiconductor sovereignty through the EU Chips Act?
Not at the frontier. The 43 billion euro European Chips Act and facilities like the Dresden TSMC fab address ten to twenty nanometre production relevant to automotive and industrial applications, but not three to seven nanometre AI-training chips. European sovereignty, as analysed in ALGORITHMUS, therefore means resilience and substitution capacity in critical sectors, not full technological autarky, which neither the fiscal envelope nor the talent base currently supports.
Claritáte in iudicio · Firmitáte in executione
For weekly analysis on capital, leadership and geopolitics: follow Dr. Raphael Nagel (LL.M.) on LinkedIn →
For weekly analysis on capital, leadership and geopolitics: follow Dr. Raphael Nagel (LL.M.) on LinkedIn →