Who Cedes?

I. The Wrong Question

When Europe lacks a capability, the first question is: where do we find the money? For the EU Chips Act it is 43 billion euros; for battery funding across two IPCEI rounds together several billion in State aid; for microelectronics, a further 8.1 billion in the second IPCEI round. Answering the money question has become the easy part; it is the only question for which there is a settled procedure.

But it is not the decisive one. In joint ventures another question decides, and it is almost never asked: Who cedes?

A consortium initially pools not capabilities but interests. Every partner has a running business it commands and a joint task it does not. Every hour a good engineer gives to the joint effort is missing from his own. And because the own carries the revenue while the joint carries none for now, every individual behaves reasonably by keeping what is safe and giving to the uncertain as little as possible.

This is neither malice nor a character flaw. It is a boardroom duty. And it adds up, in the aggregate, to a venture in which everyone participates and no one contributes.

II. What Sovereignty Means

Before one talks about resources, it should be clear what is being talked about. We propose a definition that demands neither autarky nor settles for goodwill:

Sovereign is whoever, in every relevant field, draws on outside expertise but never remains permanently dependent on it.

This is not a rejection of cooperation. Doing everything oneself would be neither possible nor sensible; the division of labour is the reason for the prosperity we are talking about. The definition prohibits not the purchase but the irreversibility.

From this follows a test one can actually apply:

What happens if the delivery stops tomorrow?

If the answer is „we will build nothing new for now“, it is a dependence in time — uncomfortable, but manageable. If the answer is „everything stops“, it is no longer a dependence but a state of being at another's mercy.

Three additions belong to the definition, or it remains a formula.

First: reversible, not merely tolerable. Reversible means one holds the knowledge to do it oneself if need be — more badly and more expensively, but still. That is the difference between buying and outsourcing. The buyer could go elsewhere. The one who has outsourced cannot.

Second: continuous practice. A capability held only on paper is no capability. Whoever has not manufactured for twenty years cannot manufacture, even with the plants and drawings in hand. Sovereignty is therefore not a stock one lays aside but a practice one maintains — and this is the real reason to start now and not when the conditions are right.

Third: the rule applies to one's own side too. Whoever demands not to remain permanently dependent cannot at the same time count on others remaining permanently dependent on him. Europe holds a unique position at one point in the semiconductor chain: ASML in Veldhoven is the only manufacturer of EUV lithography machines in the world; the optics come from Zeiss SMT in Oberkochen, the drive lasers from the environment of Trumpf in Ditzingen. Anyone who takes the definition seriously has to accept that Taiwan, Korea, and the United States are entitled to apply the same rule. A definition of sovereignty that runs in only one direction goes by another name.

This is the yardstick used in what follows.

III. How Airbus Solved It

There is one European joint venture that did not end that way, and the reason for it is constructive, not cultural.

Airbus was founded on 18 December 1970, not as a project but as a company. The legal form was the French Groupement d’Intérêt Économique (GIE) — Airbus Industrie GIE. The partners did not second personnel; they gave up plants, people, and programmes. Aérospatiale in France held 37.9 per cent, Deutsche Airbus (later DASA) another 37.9 per cent; British Aerospace joined later with 20 per cent, and CASA of Spain with 4.2 per cent. Whoever cedes cannot go on optimising on the side; the ceded capacity is no longer available on his own balance sheet.

The second constructive move is equally important and more often overlooked: the division of labour was mutually exclusive. Wings from Broughton in Wales, fuselage sections from Hamburg-Finkenwerder and Bremen, final assembly in Toulouse-Blagnac — and for none of these parts was there a second source inside the consortium. So no partner had a complete aircraft without the others, and no partner could compensate for another's exit. The cohesion lay in the construction, not in the contracts. Whether this effect was intended, or arose from the distribution of shares, cannot be settled from the accessible sources; we claim only the effect, not the intent.

Compare that with the customary form of European joint programmes: distributed work packages, each separately billable, each separately usable. That is exactly why they fall apart the moment the funding period ends. Nothing holds them together but the money, and the money runs out.

The beginning was thin enough to show the robustness of the construction. The A300 first flew on 28 October 1972 and entered service on 23 May 1974 with Air France. Then sales failed to appear: between December 1975 and May 1977 not a single new order was booked. On the apron at Toulouse stood completed but unsold aircraft — the „whitetails“. Production fell to half an aircraft per month, and the discontinuation of the programme was discussed. The breakthrough came in 1977 with Eastern Airlines: Airbus lent the customer four A300 for half a year free of charge, with only operating costs to be paid. Eastern subsequently ordered 23. By 1979, 81 A300 were in service with fourteen airlines, with 133 firm orders and 88 options. Ten years after the official launch, Airbus held 26 per cent market share by dollar value. Production of the A300 ran until 2007 and ended at 561 aircraft.

IV. The Second Case: Starting Small and Hopeless

ASML came into being in 1984 as a joint venture of Philips and ASM International in Veldhoven — as a spin-off of a project Philips no longer wanted to carry alone. The lithography-machine market in the 1980s was dominated by Nikon and Canon; ASML had no market share and no brand recognition. The first product, the PAS 2000, was a commercial and technical failure. In 1988 ASM International withdrew and Philips considered closing the venture; instead it was made independent and listed on the exchange. Only in 1991, with the PAS 5500, did the first commercial success arrive; only in 2002 did ASML become the largest supplier of photolithography systems — eighteen years after founding. The present monopoly in EUV lithography emerged another decade later still.

Two things about this are remarkable. First: the venture was bad for a long time, and they went on anyway. Second: from day one, it sold. Not demonstrators but machines that customers bought — worse ones than the competition's, at first.

This is the point at which research projects and companies part ways. The learning curve arises on the delivered unit, not on the prototype. Whoever only develops learns what might work. Whoever delivers learns what happens at a customer in continuous operation — and in manufacturing technology, that is the whole difference.

V. What Follows for a European Manufacturing Sector

The problem is usually set up like this: Europe needs a plant for leading-edge nodes, so we need billions and an operator. Then one looks for the operator, finds none, and the project is postponed — as in July 2025, when Intel cancelled its planned fab in Magdeburg. It was the flagship of the Chips Act's second pillar, estimated at some 30 billion euros; Intel's chief executive cited the absence of customer commitments.

After the foregoing, one would set it up the other way round.

First: begin at once, with what one can. Process experience exists in Europe. Infineon manufactures in Dresden, Villach, and Regensburg; STMicroelectronics in Grenoble and Catania; Bosch in Dresden and Reutlingen; GlobalFoundries in Dresden; X-FAB in Erfurt; NXP in Eindhoven; wafer suppliers such as Siltronic and Soitec; and from 2027 the ESMC in Dresden, a joint venture of TSMC with Bosch, Infineon, and NXP. The centre of gravity is at mature nodes between 22 and 90 nanometres, in power, automotive, and sensor manufacturing. What here is not top of the range is the feature size, not the discipline. Yield management, contamination control, tool availability, and statistical process control are node-independent.

Second: sell from the start. A venture that only researches never carries itself and learns too slowly. A venture that produces and sells what it can already do finances part of its own development and produces along the way exactly the experience that is needed on the way up. It grows faster and becomes independent of funding periods sooner.

Third: choose the bottleneck one can reach. Not the finest feature size. Europe's own high-performance processor SiPearl, with the Rhea chip destined for the JUPITER supercomputer at Forschungszentrum Jülich, is manufactured at TSMC — even the one notable European accelerator runs through Asian fabrication. Europe's bottleneck is not exposure but the contract manufacturing behind it. Realistic entry points lie where the competitors are less dominant: memory attachment, interconnect, advanced packaging, photonics. This is the same movement already made at the model level: not to run the race, but to look for a place where one's own means will do.

Fourth: do not keep pushing the thirty years ahead of oneself. The learning curve has a minimum duration that hangs on the number of units produced, not on the budget. One can shorten it with money but cannot skip it. Whoever waits until conditions are right will, in thirty years, have the same conditions — and a wider gap. The European Court of Auditors in its 2025 Special Report on the Chips Act found that the 20-per-cent target for 2030 will be missed; the Commission's own forecast of July 2024 puts the figure at 11.7 per cent. The cause is not a lack of resources but the structure of the approach.

VI. The Contribution Problem, More Precisely

Which leaves the question on which it actually turns: why do the partners not cede?

Because the calculation for each of them individually is unambiguous. The bread-and-butter business is known, it carries revenue, and it is controlled alone. The joint business is uncertain, slow, and must be shared. So each keeps the first and participates in the second in the hope of taking out more than he puts in.

If everyone reasons this way, what emerges is a venture with many partners and little substance. And this is the rule, not the exception.

There are two constructions that help against it, and both are uncomfortable.

One: the joint venture gets bread-and-butter business of its own. Not personnel who are seconded and can return at any time, but plants, customers, and revenue that disappear from the partners' balance sheets. That is the Airbus solution, and it hurts. That is precisely why it works.

The other: the share is set by contribution, not by pledge. Whoever brings process knowledge that demonstrably enters the line receives more; whoever comes along, less. That is co-operative logic. It requires that contributions be measured — which is costly, but less costly than a venture that falls apart after five years.

VII. Where Does It Sit on the Balance Sheet

From the foregoing follows a test question sharper than any statement of intent.

It is not: does the partner want to? It is: where does the venture sit on his balance sheet?

If it sits in the core business, it gets capacity, because the core business gets capacity anyway. If it sits alongside, it is passed over in good years for reasons of loaded capacity, because the core business is running; and it is cut in bad years, because savings must be made. There is no state of the business cycle in which a venture alongside the core business gets the rank it actually needs.

This holds between consortium partners and equally within a single house, between two business fields. And it decides independently of how good the matter is and how much money stands behind it.

Whoever plans a joint venture should therefore first check whether it sits in the core business of the partners or alongside it. If it sits alongside, only the construction from section VI helps: it needs a core business of its own.

VIII. What Would Bear Checking Before One Begins

Five questions that come before the money question.

Who cedes what, irreversibly? Not what pledges are being made, but what disappears from which balance sheet.

What does the venture sell in the first year? If the answer is „nothing“, it is a research project. That can be correct, but then it should be called that and not expected to carry itself.

Can a partner withdraw without suffering harm? If yes, he will.

Is his share set by what he pledges, or by what arrives?

Who decides when it is over? A venture without that authority does not end; it frays.

IX. Conclusion

Europe discusses sovereignty as a money question because that is the only question for which there is a procedure. For the more difficult one there is none: who cedes what he commands, so that jointly something comes into being that none can achieve alone.

Airbus solved it once, ASML needed eighteen years to market leadership and another ten to a monopoly, and both began small and badly. Whoever waits today until conditions are right will, in thirty years, be having the same debate — with the same arguments and a wider gap.

Beginning, in this case, means beginning badly and selling what one can already do.