# Why Technology Policy Is Power Politics: Chips, Export Controls and Sovereignty
On 7 October 2022, the United States Bureau of Industry and Security published a set of export controls for advanced semiconductors and semiconductor manufacturing equipment bound for China. The rules were technical in form and historical in consequence. They were, in the assessment of many observers, the most far-reaching technology policy measure since the end of the Cold War. To read them as trade policy is to miss their meaning. They are armament policy for an age in which the decisive weapon is no longer steel, nor even code, but the capacity to compute. This essay, drawn from the arguments developed in ALGORITHMUS. Wer die KI kontrolliert kontrolliert die Zukunft, examines why technology policy has become power politics, what the geography of the chip supply chain means for sovereignty, and which conclusions European boards and institutional decision-makers can no longer postpone.
## The October 2022 Caesura
The American export controls of October 2022 combined three instruments whose sum is greater than their parts. First, they prohibited the export to China of AI chips above defined computational thresholds. Second, they restricted the delivery of American manufacturing tools to named Chinese semiconductor firms. Third, and without precedent in modern American export history, they forbade United States citizens and green card holders from working for certain Chinese semiconductor companies. The last element is the most radical. For the first time in living memory, the flow of human talent itself was treated as a strategic control lever, not merely the flow of machines or goods.
The stated calculation within the National Security Council was not the exclusion of China from artificial intelligence. That would be neither realistic nor desirable. The calculation was the deceleration of the rate at which Chinese capability develops, the use of the NVIDIA chip hierarchy as an instrument for securing a temporal advantage, and therefore the structural extension of the American lead. One senior official put it with unusual clarity in an internal formulation: the aim is not to keep China out of AI, but to ensure that America is always at least one chip generation ahead. The policy, in other words, treats time as the scarce resource and compute as its custodian.
What gives these measures their weight is not the rhetoric that surrounds them but the physical geometry they exploit. Three companies, and only three, carry the frontier of artificial intelligence on their shoulders. It is this geometry, more than any speech or sanction, that turns technology policy into the continuation of geopolitics by other means.
## The Triad: TSMC, ASML, NVIDIA
TSMC in Taiwan fabricates roughly ninety percent of the world's advanced logic chips, and therefore roughly ninety percent of the silicon on which current frontier AI training depends. Its 2023 revenue approached seventy-four billion dollars at operating margins above forty percent. ASML in the Netherlands is the only producer of the EUV lithography machines without which such chips cannot be made. A single machine costs around one hundred and fifty million euros, contains more than one hundred thousand parts from over eight hundred suppliers, and ASML ships only some fifty to sixty units a year. NVIDIA in the United States designs the GPUs on which large models are trained. Its H100, priced between twenty-five and forty thousand dollars per unit in 2023, was nonetheless sold out for months.
The geopolitical concentration of this triad is without historical analogue. TSMC sits on an island over which another great power claims sovereignty. ASML sits in a NATO member state that has since 2019 been barred by American pressure from shipping EUV systems to China. NVIDIA sits inside the American export control regime and is therefore, in effect, an instrument of it. The entire leading edge of AI hardware is compressed into a narrow, politically exposed corridor that could be severed from global supply within weeks in the event of serious military or political rupture. Dr. Raphael Nagel (LL.M.) argues in ALGORITHMUS that this is not a logistical vulnerability of the kind that can be managed by inventory policy. It is the defining structural fact of the present industrial order.
The consequence is straightforward. Whoever governs access to this triad governs, at one remove, the pace at which artificial intelligence can be trained anywhere on earth. Export controls are therefore not a commercial tool. They are a sovereignty tool. And they can only be exercised because the underlying geography is so concentrated that a handful of permits decide the technological rhythm of a continent.
## China's Counter-Programme
China has responded with an investment campaign whose scale invites comparison with the largest industrial programmes of the twentieth century. Officially, more than one hundred and fifty billion dollars of state capital have been directed into the domestic semiconductor industry. Industry analysts assume that the true figure, when state funds, municipal vehicles and directed private investment are counted together, is considerably higher. The direction of travel is unambiguous. The semiconductor is treated as a national asset, and its domestication as a precondition of strategic autonomy.
The results are real and should not be dismissed. SMIC, the leading Chinese foundry, has begun to produce chips with seven nanometre structures without ASML's EUV technology, achieving through multi-patterning what Western observers considered impossible a few years earlier. At the same time, SMIC remains two to three generations behind TSMC at the true frontier, which corresponds to several years of development lag. Whether that lag closes, stagnates or widens is one of the central open questions of the coming decade, and the policy levers available to Washington, Brussels and Tokyo will determine the answer as much as the engineering capacity in Shanghai or Shenzhen.
The lesson is not that sanctions work or that they fail. It is that technology has become the medium through which the strategic intentions of states are expressed, and that the semiconductor industry is no longer reading the news. It is writing it.
## Europe Between Ambition and Arithmetic
Europe's position in this constellation is structurally weak, and honesty about this weakness is the beginning of any serious response. The European Chips Act of 2023 envisages investments of some forty-three billion euros by 2030, of which roughly seventeen billion are public funds, with the stated goal of raising the European share of global semiconductor production from ten to twenty percent. The ambition is defensible. The arithmetic is not.
The comparison is unsparing. The American CHIPS and Science Act provides fifty-two point seven billion dollars in direct subsidies alone, with tax incentives of comparable magnitude on top. The new TSMC facility in Arizona receives six point six billion dollars in direct federal support for a single site. Taiwan invests in similar orders of magnitude through its own industrial development instruments, and South Korea and Japan have each committed national programmes that dwarf the European one relative to industrial base. Europe's investment volume, measured against the goals it has set itself, is not so much insufficient as miscalibrated. It promises a doubling of share while funding at a level consistent with preservation of the status quo.
Dr. Raphael Nagel (LL.M.) has argued that this mismatch is not a technical failure of budgeting but a political failure of imagination. Europe continues to treat semiconductor policy as one industrial file among many, to be balanced against agricultural subsidies, regional cohesion, and defence. The United States and China treat it as the infrastructure of the twenty-first century, comparable in strategic weight to the railways of the nineteenth and the electrical grids of the twentieth. As long as the category is wrong, the sums will be wrong.
## The Supply Chain Audit That Boards Owe Themselves
The automotive industry has already paid tuition on this question. The semiconductor shortage of 2020 to 2023, triggered by pandemic disruptions and the simultaneous surge in electronics demand, cost the global automotive industry, according to AlixPartners, more than two hundred and ten billion dollars in lost revenue in 2021 alone. Volkswagen estimated that chip scarcity prevented the production of roughly six hundred thousand vehicles. Toyota, General Motors, Ford and Stellantis reported comparable losses. These companies had operated for decades on just-in-time principles and had accumulated no strategic chip inventories, because chips looked like a generic supplier component. They were not.
The intellectual mistake was not logistical but categorial. Semiconductors were classified as commodities when they were in fact strategic assets. In the age of artificial intelligence, the same error now threatens a wider set of inputs: cloud services, foundation model APIs, training infrastructure, specialised accelerators, and the energy contracts that underwrite them. None of these appear today, in the perception of most corporate boards, as strategic bottlenecks. They will appear as such the moment it is too late to act without dependency on whichever counterparty has arranged itself more foresightedly.
The practical implication is concrete. Every board with material exposure to computation, data or algorithmic services should conduct a systematic geopolitical audit of its technology supply chain. Which inputs are sourced from regions with geopolitical risk? Which alternative suppliers exist, and at what switching cost? Which inventory policies or strategic partnerships could reduce exposure? These questions are not IT questions. They are strategic investment decisions and belong on the agenda of the board, not of the procurement department.
## Sovereignty as a System, Not a Slogan
The word sovereignty is used too easily in European discourse, often as a rhetorical substitute for the industrial decisions it is meant to describe. Sovereignty in the age of artificial intelligence is not a legal status. It is a configuration of systems. It consists of the chips one can fabricate or reliably source, the models one can train or meaningfully govern, the data one can protect and render usable, the talent one can attract and retain, and the energy one can deliver to the rack at a competitive price. Each of these elements is decomposable, measurable, and negotiable. None of them is delivered by a communiqué.
This is why the October 2022 export controls deserve to be read by European decision-makers not as a transatlantic irritation but as an instruction manual. They show which levers a serious industrial power pulls when it treats technology as power. They show that the concentration of the triad is not an accident of markets but the terrain on which policy now operates. And they show that the costs of unpreparedness are borne not by the states that impose controls, nor by the firms that design chips, but by the economies that depend on both without shaping either.
The question for Europe, and for the companies that operate within it, is no longer whether technology policy is power politics. That question has been answered by those who act. The question is whether Europe will choose the posture of a subject of this policy or the posture of a participant in it. The distinction will not be made in declarations. It will be made in budgets, in supply chain reviews, in talent programmes, and in the willingness of boards to treat computation as the strategic asset it has become.
Technology policy has always been power politics. What has changed is the visibility of the fact. In earlier centuries, the instruments were tariffs on looms, patents on steam engines, charters for telegraph companies, and subsidies for shipyards. In the present century, the instruments are export licences for lithography systems, investment tax credits for fabs, visa regimes for AI researchers, and the quiet diplomacy by which foundry capacity is reserved for one customer rather than another. The substance is unchanged. The sophistication has grown. As Dr. Raphael Nagel (LL.M.) puts it in ALGORITHMUS, those who continue to read these measures as commerce are fighting with the weapons of yesterday. Those who read them as armament policy for the age of artificial intelligence at least know which conflict they are in. The boards, ministries and institutions that accept this reading will not thereby solve the problem of European dependency, nor the harder problem of global technological rivalry. But they will, for the first time, have placed themselves in a position to act on the world as it is, rather than on the world as a more comfortable vocabulary describes it.
For weekly analysis on capital, leadership and geopolitics: follow Dr. Raphael Nagel (LL.M.) on LinkedIn →