The Question We Taught the World to Ask, and Then Refused to Fund

In 1948 mat een Nederlander de energie van het lege niets. In 2026 haalde een Amerikaan twaalf miljoen dollar op om dat lege niets aan ons terug te verkopen. Dit stuk gaat over de vraag of we déze keer wél op tijd zijn — en over het agentschap dat nu wordt opgericht, en daar precies voor bedoeld is.

(The argument below is in English, for the international partners it also needs to reach.)

An open argument to the cabinet, and to anyone who will be asked to pay for it. A confidential project plan, with ownership structure, security regime, and partner terms, exists for those in a position to act on it; an invitation to request it appears at the end.

In 1948, in a laboratory in Eindhoven, a Dutch physicist named Hendrik Casimir was helping a colleague with a mundane industrial puzzle: why suspensions of fine powder, the kind Philips used in making its lamps and television tubes, did not settle the way the theory of the day said they should. He ended up proving something far stranger than the question deserved: that empty space is not empty, that it presses on things, that there is energy in the void itself. He worked it out, and later it was measured, again and again, until no serious physicist doubted it. It is one of the most consequential Dutch discoveries you have never heard of.

Seventy-eight years later, in the spring of 2026, a former NASA engineer in Texas raised twelve million dollars to sell it back to us.

This is a story about a question the Netherlands taught the world to ask, and then left lying on the table for the better part of a century. It is also a story about an agency the country is deciding, right now, whether to build, and which was conceived for exactly this kind of unfinished business.

The Netherlands is on the verge of doing something it has never done before. In the coalition agreement of 30 January 2026, the incoming D66–VVD–CDA government reserved a €500 million capital injection for a National Agency for Disruptive Innovation, NADI, modelled on the American DARPA, the British ARIA, and the German SPRIN-D. The agency is not yet built — the coalition agreement calls it the "agency still to be established," a minority government must still win majorities for it measure by measure, and the Instituut voor Publieke Economie has warned that the money as reserved may not be the right kind. But the direction is set, the analysis is done, and the decisions that will shape it are being taken this year. The idea, set out in Peter Wennink's report of late 2025 and assessed in economic affairs minister Karremans's Kamerbrief of February 2026, is simple to state and hard to do: give a small number of talented programme directors real money, real autonomy, and a high tolerance for failure, and let them chase breakthroughs that the normal machinery of Dutch science and industry is too cautious, too slow, or too fragmented to attempt.

NADI will live or die by the programmes it chooses. An agency built to take bold bets that funds only safe ones is a contradiction that will collapse under its own timidity, and its critics, who already warn it risks becoming just another department issuing just another kind of subsidy, will be proven right. So the question of which programmes NADI runs first is not a detail. It is the whole test.

This is a proposal for one of them, and it is a detailed proposal, not a proefballon. I am going to argue that NADI should fund a serious, properly resourced, Dutch-led effort to answer a question mainstream physics has spent seventy years assuming is closed, that no one is yet testing with the rigour it deserves. The question is whether usable energy can be extracted from the quantum vacuum, the energy that exists in empty space. And before I make the case, let me put my own credibility on the table: I think it most likely fails. If you made me bet, I would bet against net energy ever coming out of the vacuum. I am going to ask you to fund it anyway, and the rest of this piece is the argument for why that is not a contradiction but the whole point of what NADI is for.

The downside of not trying, and being wrong about that, is a story this country already knows by heart. The World Wide Web was invented in 1989 in a European physics laboratory, by a European scientist, on public money. Europe then spent thirty years watching the trillion-euro businesses built on top of it, the search engines, the marketplaces, the social networks, grow up somewhere else entirely. We wrote the protocol. Others wrote the invoices. It would be a peculiar kind of failure to do that twice with a technology we can trace to a laboratory bench in Eindhoven.

I want to be honest about the timing, because it is the reason this essay has urgency it did not have when I began drafting it. While I was writing, in the spring of 2026, an American startup exited stealth on precisely this idea. I will come to them in detail, because they matter, and because what they are and are not changes the argument in the Netherlands' favour rather than against it. But let me build the case first, and begin not with the vacuum, but with a pattern.

Part I: The pattern

There is a particular kind of statement that physicists learn to distrust, and it sounds like this: "That is impossible, and here is the equation that proves it."

Not because the equations are wrong. They are usually correct, given their assumptions. The trouble is always the assumptions, the quiet premises sitting underneath the mathematics that nobody thought to question because they seemed too obvious to state. When one of those premises turns out to be incomplete, the "impossible" becomes a Tuesday afternoon in a laboratory.

What follows are four stories. I have chosen four deliberately, not one, and not two. A single example proves nothing; anyone can find a lucky exception and wave it around. Two might be coincidence. But four cases, drawn from four different fields, across four different generations, each following the same arc from confident denial to working demonstration, begin to look less like a collection of flukes and more like a feature of how knowledge actually advances. The pattern is the point.

One: The men who proved flight was impossible, shortly before it happened

In the 1890s, Lord Kelvin was the most authoritative scientific voice in the English-speaking world. He had helped lay the first transatlantic telegraph cable, his temperature scale is still the one physicists use, and when he spoke about what nature would and would not permit, people listened. In 1895 he pronounced heavier-than-air flying machines impossible, and the following year, declining an invitation to join the Aeronautical Society, he wrote that he had "not the smallest molecule of faith in aerial navigation other than ballooning."

He was not alone, and he was not being foolish. The skepticism rested on real calculations. The astronomer Simon Newcomb, one of America's most respected scientists, argued around the turn of the century that the power required to keep a heavy machine aloft scaled badly against the power any engine could produce. Birds, the reasoning went, sat at the very edge of what the physics allowed, and a machine made of wood, metal, and a heavy motor could not possibly cross that line. The math on power-to-weight ratios was not nonsense. It was, given the assumptions of the day about wings and engines, fairly persuasive.

The assumptions were wrong. Two bicycle mechanics in Ohio, who had no standing in physics whatsoever, did something the theoreticians had not: they built a wind tunnel, measured lift and drag empirically, discovered that the published aerodynamic coefficients everyone relied on were inaccurate, and redesigned their wings around their own data. On 17 December 1903, the Wright Flyer left the ground at Kitty Hawk under its own power and stayed there.

From Kelvin's pronouncement to powered flight: roughly eight years.

Two: From "moonshine" to a reactor under a football stadium, in nine years

On 11 September 1933, Ernest Rutherford, the man who had discovered the structure of the atom, addressed the British Association. Asked about the prospect of drawing usable power from atomic nuclei, he was blunt. Anyone who looked for a source of power in the transformation of atoms, he said, was talking "moonshine."

Again, this was not ignorance. Rutherford understood the atom better than any living person. The problem, as the physics of 1933 understood it, was that prying energy out of a nucleus cost more than the nucleus gave back. You had to fight the forces holding it together, and the books did not balance.

The Hungarian physicist Leó Szilárd read Rutherford's remark in the newspaper and was irritated by it. The story goes that that same morning, waiting at a traffic light on Southampton Row in London, he worked out the idea that demolished Rutherford's assumption: a nuclear chain reaction. If splitting one atom released neutrons, and those neutrons split further atoms, releasing more neutrons, then you would not have to pay the entry cost over and over. The reaction would feed itself. The concept simply did not exist when Rutherford called the whole thing moonshine, so naturally his calculation had not included it.

Nine years later, on 2 December 1942, a team led by Enrico Fermi assembled a pile of graphite and uranium in a squash court beneath the stands of the University of Chicago's football stadium and produced the first self-sustaining nuclear chain reaction in human history. The "moonshine" was now a controlled reactor humming in the middle of a major city.

From "moonshine" to the first working nuclear reactor: nine years.

Three: The temperature ceiling that physics could prove, until it could not

This one is subtler, because the theory involved was genuinely excellent.

In 1957, three physicists produced what is still considered one of the great triumphs of twentieth-century physics: the BCS theory of superconductivity, which finally explained why certain metals lose all electrical resistance when cooled to near absolute zero. The theory worked beautifully. It explained the existing data, predicted new effects, and won its authors the Nobel Prize. And it carried an implication that hardened into common knowledge: superconductivity was a phenomenon of extreme cold. The delicate pairs of electrons that carried the resistance-free current would be shaken apart by heat well below around thirty degrees above absolute zero. Above that, the textbooks taught, superconductivity could not survive.

For nearly three decades, this was simply understood to be how the world worked. Then, in 1986, two researchers at IBM's Zurich laboratory, Georg Bednorz and Alex Müller, started testing a class of brittle ceramic oxides, materials that, being closer to the stuff of coffee mugs than of wires, nobody expected to superconduct at all. They found superconductivity at thirty-five degrees, brushing against the supposed ceiling. Within a year, other groups pushed the same family of materials past ninety degrees above absolute zero, comfortably above the temperature of liquid nitrogen, which transformed superconductivity from an exotic laboratory condition into something achievable with a cheap and abundant coolant.

The Nobel committee, which usually waits decades to be sure, awarded Bednorz and Müller the prize in 1987, the year after their discovery. And here is the part that should stay with you: nearly forty years later, we still do not have a complete, agreed theory of why these high-temperature superconductors work. The barrier was real in the sense that the old theory really did forbid it. The barrier was false in the sense that the old theory was not the whole story.

Four: Extracting order from chaos, which the Second Law forbids and your cells do constantly

The last of the four is the one that matters most for what I am going to argue, because the "impossible" here is the most fundamental of all: the Second Law of Thermodynamics.

The Second Law is as close to sacred as physics gets. One of its consequences is that you cannot extract useful, directed work from random thermal jiggling at equilibrium. The molecules of the air in your room are battering everything from all sides, carrying a great deal of energy, but because the battering is symmetric, random in every direction, you cannot harness it to turn a wheel. Richard Feynman made this rigorous in a famous analysis of a tiny ratchet and pawl, a microscopic mechanism that looks like it should let random molecular collisions turn it one way only. He showed, with care, that at equilibrium it cannot work. The same randomness that would push the ratchet forward also knocks the pawl loose to let it slip back. The books balance, and you get nothing. Maxwell's celebrated "demon," a hypothetical being that sorts fast molecules from slow to create useful order, fails for the same deep reason.

And yet.

Inside every cell of your body, right now, molecular machines are extracting directed motion from exactly that random thermal storm. Kinesin molecules walk in a straight line along microscopic tracks, hauling cargo, taking step after deliberate step, powered in part by the same Brownian buffeting that Feynman proved could do no work. The enzyme that makes ATP, the fuel of all life, is a literal rotary motor turned by molecular chaos. Biology has been running trillions of these impossible little engines in every living thing for billions of years.

The resolution is beautiful, and it is the key to everything that follows. These motors do not violate the Second Law. They cheat the assumption hidden inside the simple version of it, the assumption of equilibrium. The cell is not at equilibrium. It maintains a chemical gradient, burning ATP, and the molecular motor couples to that gradient. The random thermal motion is not the fuel; it is the transmission. The asymmetric structure of the motor, combined with an energy source that keeps the system away from equilibrium, turns undirected noise into directed work.

The lesson is not that thermodynamics is wrong. Thermodynamics is never wrong. The lesson is that "you cannot get directed work from random fluctuations" quietly assumed equilibrium, and the moment you find a system that is not at equilibrium, the door that was bolted shut swings open.

Why I told you four stories instead of one

If I had told you only the flight story, you could fairly dismiss it. Engineers are not physicists; perhaps Kelvin was simply outside his domain. If I had added only the nuclear story, you might say that early-twentieth-century physics was young and prone to error. The reason I insisted on four, spanning aerodynamics, nuclear physics, condensed-matter physics, and thermodynamics, from the 1890s to the present, is that the repetition is the evidence.

Each of these "impossibilities" was pronounced by serious people, backed by real mathematics, and believed by the mainstream of an entire field. Each was not wrong in its mathematics but incomplete in its premises. And each gave way, often within a decade, to a working demonstration, sometimes built by outsiders, sometimes resting on physics we still do not fully understand. This is not a story about a few embarrassing predictions that smart people happened to get wrong. It is a description of the normal texture of scientific progress. The frontier is, by definition, the place where our current theories run out, and our current theories are precisely the thing that tells us what is impossible.

So when someone shows me an equation proving that a thing cannot be done, I no longer ask, "Is the equation correct?" It usually is. I ask, "What is this equation quietly assuming, and how confident are we in that?"

Part II: And then there is solar

Before I turn to the vacuum, one more story, because it teaches a different lesson than the four, and that lesson is just as important to what follows.

The four above are tales of "impossible" overturned. Solar power is not quite that. The physics behind it, the photovoltaic effect, the fact that light striking certain materials knocks loose a current, was discovered by Edmond Becquerel in 1839 and was never thought impossible. It was thought useless. For more than a century it was a laboratory curiosity with no serious prospect as a source of power. When Bell Labs built the first practical silicon solar cell in 1954, it converted about six percent of sunlight to electricity at a price so absurd that the only customer who could justify it was a satellite, where there was no alternative at any price. Serious, well-informed people looked at solar in 1960, or 1980, and concluded, with sound arithmetic, that it would remain a niche curiosity forever, far too expensive to ever matter for the grid.

They were wrong, and they were wrong in a way nobody saw coming. There was no single breakthrough, no eureka. Instead, decade after decade, manufacturing scale and a thousand incremental improvements drove the cost down along a relentless curve, until the International Energy Agency reported in 2020 that solar had become, under favourable conditions, the cheapest source of electricity in history. The transformation did not come from new fundamental physics. It came from the application of an old, dismissed effect, scaled and engineered until the economics inverted. Almost no one predicted the magnitude of it, and almost no one predicted the timing.

Two lessons sit inside that story, and I will use both. The first is that the road from "interesting curiosity" to "cheapest power on Earth" can run for a century and turn on an unforeseen accumulation of engineering, which should make us humble about declaring any real, measured effect permanently useless. The second, and for the hard-nosed reader the more important, is that the revolution can live in the applications rather than the original effect. Hold that second lesson. It is the foundation of the argument, later, that even if the vacuum programme fails at its grandest goal, the engineering it produces along the way may be worth more than the programme costs.

Part III: The question, and the people finally asking it

Empty space is not empty. This is not speculation; it is among the best-tested facts in physics, and the story of how we learned it begins, as it happens, in the Netherlands.

In 1948, the Dutch physicist Hendrik Casimir, working at the Philips laboratories in Eindhoven, was drawn into a thoroughly practical industrial problem: a colleague's measurements of the forces between particles in colloidal suspensions did not match the accepted theory. Out of that question came one of the strangest predictions of modern physics. Casimir calculated that two metal plates, placed extremely close together in a perfect vacuum, with all air and all atoms removed, would nonetheless feel a force pushing them together. The push came from the vacuum itself. According to quantum theory, the vacuum is not a still, dead nothing; it seethes with fluctuating fields, a restless background of energy present even when everything that could be removed has been removed. Between two close plates, fewer of those fluctuations fit than outside, so the outside presses harder than the inside, and the plates are shoved together. Empty space, Casimir was saying, pushes.

For a long time this was a prediction more admired than confirmed, because the force is fantastically small and fantastically hard to measure. The first serious experimental attempt was made in 1958 by Marcus Sparnaay, also at Philips in Eindhoven, and his results were consistent with Casimir but too crude to count as proof. The clean confirmation took another four decades. In 1997 the physicist Steve Lamoreaux, using a delicate torsion pendulum, measured the Casimir force to within a few percent of the prediction. Since then it has been confirmed again and again, in laboratory after laboratory, to ever finer precision. There is now no serious doubt about any of it. The energy of the vacuum is as real and as measured as the weight of a stone.

Hold onto that, because it is the foundation of the whole argument: we are not asking whether vacuum energy exists. We know it exists. We have known since a Dutchman measured it, and confirmed it since with some of the most careful experiments physics has produced. The only open question is what, if anything, we can do with it.

From measuring the vacuum to commanding it

For most of the twentieth century the answer to "what can we do with it" was "nothing; we can only observe it." That has changed, quietly and recently, and almost nobody outside the field has noticed. In the last fifteen years the work has moved from measuring the vacuum to manipulating it, and the pace has been accelerating.

Start with the assumption that the Casimir force only ever pulls. For half a century that was taken as given. Then in 2009 a pair of physicists, Jeremy Munday and Federico Capasso, working at Harvard, measured a repulsive Casimir force, the vacuum holding two surfaces apart rather than dragging them together, by choosing the right materials and immersing the gap in a carefully chosen fluid. The single fixed behaviour everyone had built their intuitions around turned out to be one setting on a dial.

Then came the result that should be far more famous than it is. In 2011, a team at Chalmers University in Sweden, led by Per Delsing with Christopher Wilson, did something that sounds like a conjuring trick and is in fact careful quantum engineering. They built a "mirror" out of superconducting circuitry whose effective position could be changed electronically, and they wiggled it at a few percent of the speed of light, far faster than any physical mirror could ever move. When you shake the boundary of the vacuum that violently, theory predicts that some of the fluctuating, "virtual" energy of empty space gets promoted into real, permanent, detectable particles of light. The Chalmers group detected exactly that: a faint glow of microwave photons, conjured out of the vacuum by nothing but a rapidly changing boundary. They made light from nothing.

I want to be careful here, because this is precisely where honest argument and wishful thinking part ways. That experiment did not produce surplus energy. It cost more to wiggle the mirror than the harvested photons carried away; the books still balanced. But notice the resemblance to the molecular motors in your cells. There, an energy input plus an asymmetric structure converts random thermal noise into directed, useful work. Here, an energy input plus a rapidly changing boundary converts the random fluctuations of the vacuum into real, countable quanta. The Chalmers result proves the vacuum's restless energy is not forever locked away as an untouchable background. Under the right conditions it can be coaxed across the line into something real and measurable. Whether any configuration can do so at a net profit is the open question. But the door, long believed welded shut, has visibly moved on its hinges.

And the manipulation has only grown more sophisticated. In 2024, researchers at the University of Science and Technology of China suspended a tiny gold sphere over a silica plate in a water-based fluid carrying magnetic nanoparticles, and showed that by varying an external magnetic field they could tune the Casimir force between attraction and repulsion. The same year, Jeremy Munday's group, now at the University of California, published a method for controlling the force with biased semiconductors and an applied voltage, and in 2025 his lab reported experimentally tuning it with engineered three-dimensional nanostructures. Some of this is theory, some experiment; some is already contested and being argued over in the journals, which is exactly what a live field looks like. Taken together, these results describe a field that has crossed a threshold. The Casimir effect is no longer a fixed quirk of nature we can only watch. It is becoming a knob we can turn.

The most embarrassing number in physics

Now set against all this control a single humiliating fact.

When physicists use quantum theory to calculate how much energy the vacuum ought to contain, and compare that figure to how much it evidently contains, judged by its gravitational effect on the expansion of the universe, the two numbers disagree. They do not disagree by a little. They disagree by a factor of ten followed by roughly a hundred and twenty zeros. It is, by an enormous margin, the worst quantitative prediction in the history of science, so bad that physicists have given it a name with a whiff of gallows humour: the vacuum catastrophe.

This matters to the argument more than any single experiment, so let me be plain about what it means. It means that the thing we are talking about manipulating, the energy of empty space, is something our best theories describe with breathtaking inaccuracy. We do not have a small, tidy gap in our understanding of the vacuum, the kind that gets filled in by a graduate student over a productive summer. We have a chasm. And it is into exactly this chasm that the confident statement "you can never extract energy from the vacuum, because it is the ground state" is dropped, as though it were settled. The "ground state" argument is the vacuum's version of the equilibrium assumption from the molecular-motor story. It is the quiet premise underneath the mathematics. And it is being asserted about a quantity our theories misjudge by a hundred and twenty orders of magnitude.

The Americans have already started

Here is where the timing turns urgent, and where I have to be scrupulously fair, because the facts cut in two directions at once.

For years the energy question, whether the vacuum can be made to yield net power rather than merely be nudged, was chased almost entirely by lone figures on the margins. The most persistent is Garret Moddel, a professor at the University of Colorado, who has built small layered devices incorporating Casimir cavities that he reports produce tiny, unexplained amounts of electrical power, apparently drawn from the vacuum. He has published in peer-reviewed journals and filed patents. His results are, in his own careful words, tantalising but inconclusive: small, hard to rule clean of measurement artefacts, not the unambiguous demonstration that would force the field to take notice.

The one institution that took the question seriously with real money was, tellingly, the research arm of the United States military. Through a programme first called QUEST and then ARRIVE, an acronym for Applications Resulting from Recent Insights in Vacuum Engineering, the American defence agency DARPA funded exactly this line of work, in its own words testing whether the energy density of the quantum vacuum can be controlled and transduced into a usable form. That programme has now run its course. And out of the American vacuum-research ecosystem it helped seed, in the spring of 2026, a company emerged from stealth that a Dutch policymaker needs to know about.

The company is called Casimir Inc. It was founded by Harold "Sonny" White, formerly of NASA's Eagleworks advanced-propulsion laboratory, and it raised a twelve-million-dollar seed round led by Scout Ventures, with Tim Draper among the backers; the technology was incubated at the space entrepreneur Kam Ghaffarian's Limitless Space Institute. It claims a chip, "MicroSparc," built from arrays of fixed Casimir cavities, that draws a continuous trickle of electrical power from vacuum fluctuations by way of a mechanism its founder describes, in words that will sound familiar, as "a kind of quantum ratchet." It targets commercial devices by 2028. It rests on a theoretical paper, "Emergent Quantization from a Dynamic Vacuum," that its founder published in Physical Review Research in March 2026.

Now the honest part. You should treat Casimir Inc.'s claims with heavy caution, and I do. White's scientific track record is contested: he led the group behind the "EmDrive" thruster whose reported thrust was widely attributed to measurement error and, when tested independently at TU Dresden, disappeared under proper controls. His theoretical framing of the quantum vacuum has been challenged by mainstream physicists for years. The MicroSparc result is self-reported, not independently replicated, a trickle of microamps rather than a demonstrated, audited net-energy device, and several physicists approached by reporters declined even to comment, while others filed the claim alongside the long history of "free energy" announcements that dissolve on inspection. I am not citing Casimir Inc. as proof that vacuum energy extraction works. It is not proof of anything yet.

Let me be fair to the skeptics before I use them, because most physicists who doubt this are not being timid or foolish; they are applying a sound prior built on a century of "free energy" claims that turned out to be measurement errors or worse, and that prior has served the field well. The argument here is not that they are wrong to be skeptical. It is that skepticism is a reason to test carefully, not a reason to refuse to look, especially once serious money has entered the field.

What Casimir Inc. is, then, is a signal, and the signal is threefold. First, serious American capital and a serious American defence agency have decided this question is worth real money, which retires the easy dismissal that only cranks take it seriously. Second, the field is now moving fast enough that a company will make grand claims before anyone has settled the underlying physics, which means the ground is about to be contested by a mixture of genuine science and Silicon-Valley overclaim, with no rigorous, independent, publicly accountable referee anywhere in sight. And third, the specific mechanism being commercialised, the fixed-geometry Casimir ratchet, is the same one this essay argues the Netherlands is uniquely equipped to investigate properly. The Americans have not answered the question. They have started asking it loudly, with money, and with a credibility problem. That is the gap.

What I am, and am not, claiming

Let me be exact, because this is the precise point where serious inquiry is mistaken for fantasy.

I am not claiming that energy can be extracted from the vacuum. The honest position is that it most likely cannot; the mainstream is more often right than wrong. If pressed to put rough numbers on my own view, and they are my own subjective estimates, not a calculation, I would say the chance that net extraction is simply impossible is high, perhaps three in four; the chance that some loophole exists but yields too little to matter, meaningful but smaller; and the chance of a real, usefully large effect, genuinely small, a few percent. I would not fund this on a promise. I would fund it on that few percent, because of what sits behind it.

What I am claiming is narrower, and I think undeniable. We are staring at a phenomenon, the energy of the vacuum, that we have measured beyond doubt and now demonstrably know how to manipulate. We do not understand it; our best theories misjudge its magnitude more severely than they misjudge anything else in all of science. The case that we can never exploit it rests on an assumption of exactly the type history keeps overturning. Serious money is now chasing it. And nobody, anywhere, is investigating it with the rigour, independence, and fabrication strength required to give the world an answer it can trust. Those facts, stacked, do not describe a closed question. They describe an open one that is about to be defined by others, badly, unless someone does it properly.

NADI was created for questions shaped exactly like this one.

Part IV: Why this is a NADI programme, and what it would actually look like

It is one thing to say a question is worth asking. It is another to show that there is a real plan, run by real institutions, with real numbers, real off-ramps, a real ownership structure, and a real security regime, rather than a daydream with a budget attached. The rest of this argument is the plan, because the difference between a moonshot and a fantasy is entirely in the engineering of the bet.

The fit with NADI is almost uncomfortably precise

NADI's designers did not have the vacuum in mind, but they could hardly have described this programme more exactly if they had.

NADI is meant to fund high-risk, high-reward work the ordinary system will not touch, judged at the level of the portfolio rather than the individual project, so that inevitable failures are expected and the rare success pays for the rest. This is the textbook case: a question with a small probability of an enormous payoff, unfundable through normal channels precisely because it sits next to a century of cranks and frightens every grant committee that sees it.

NADI runs its programmes through autonomous programme directors who manage competing teams and release funding only as hard milestones are met, milestones designed so that clearing them means a real breakthrough has been realised. That is exactly the structure below: parallel groups in the Netherlands and Sweden attacking different geometries, money released in tranches at pre-defined go/no-go gates, each gate a genuine scientific question with a yes or no answer.

NADI's own design holds each programme to a demanding bar: to be judged successful at the finish, a programme should generate on the order of seven to ten times its cost in societal spillover. I will show this programme clears that bar even in the scenarios where the physics fails, and obliterates it in the scenario where it succeeds, which very few candidate programmes can honestly claim.

And two of the four strategic themes proposed as NADI's starting ground, energy-and-climate technology and security-and-resilience, are exactly where this programme sits. It is not a fringe fit. It is squarely inside the agency's stated first priorities.

There is a fair objection here that deserves answering head-on, because it is the one a thoughtful official will actually raise. NADI will have only a handful of founding slots, and there are worthier-sounding candidates queuing for them: quantum-safe cryptography, the protection of undersea infrastructure, grid-scale energy storage. Why spend one of those scarce slots on a bet that will probably fail? The answer is not that this programme is safer than those. It is that a portfolio composed entirely of sensible, likely-to-deliver programmes is not an ARPA at all; it is a well-run subsidy scheme, and it will fail at the one thing NADI was created to do, which is to take the swings that the ordinary system cannot. Every genuine ARPA portfolio needs its highest-variance bet, the one that looks slightly mad and would be transformative if it landed. The question is not whether NADI should hold one such bet — it must, or it has misunderstood itself — but which candidate is the best version of it. This one has an unusually strong claim: the physics is Dutch in origin, the fabrication and metrology to pursue it are already here, the competition has just started moving, and the failure case still pays. If not this, then some other moonshot must fill that slot; my argument is that few available moonshots combine so much national advantage with so large a prize.

There is even a fit most readers will miss, and it cuts to a real tension in NADI's own funding. The €500 million the coalition set aside is structured as a capital injection, attractive politically because it does not score against the deficit, but awkward for pure grant-giving, since money expected to earn a return is not the same as money spent on risky research. The clean resolution, which this programme is built around, is to keep the two kinds of public money in separate instruments: NADI funds the research as a grant (which honestly scores as expenditure, the correct treatment for ARPA-style risk) and receives national-security control rights in return, while a separate public investor, InvestNL, may take returnable equity alongside private capital. The structure I propose does not merely fit NADI's mission; it fits, and respects, the specific mechanics through which NADI has been funded.

The shape of the programme: three phases, two gates, no blank cheques

The programme is built so that we are never, at any moment, betting everything on the final answer. It proceeds in three phases, with two hard decision gates between them. At each gate the programme either earns its next tranche with evidence, or it stops, and stopping still leaves the country with something valuable in hand.

Phase 1, "does the vacuum lean?", years one to three, about €30 million. Can an asymmetric geometry produce any directional effect at all? We build, in the Netherlands, the capability to run dynamic-vacuum experiments that until now only a handful of labs can do, in partnership with the Swedish group that invented them. We design and fabricate asymmetric Casimir structures, the nanoscale equivalent of ratchet teeth, and measure, with rigorous energy accounting, whether the vacuum pushes through them unevenly. The gate: a directional effect, reproduced independently by more than one laboratory. If we cannot show it, the programme stops, having spent a sum comparable to a motorway interchange to retire a serious question and build a world-class fabrication and measurement capability that does not evaporate.

Phase 2, "does it profit?", years three to six, about €50 million. Only if Phase 1 passes: does any configuration yield a genuine net surplus of energy over a complete cycle, once every input is honestly counted? This is where the most careful calorimetry lives, and where independent replication is mandatory, because this is the phase where it would be easiest to fool ourselves, and where a self-reported trickle of microamps, the level of today's claims, would be nowhere near good enough. The gate: net positive energy, confirmed by laboratories that did not build the device. If we cannot show it, the programme stops again, and redirects everything built toward the very real commercial applications of Casimir-force control.

Phase 3, "can we build it?", years six to ten, about €70 million. Only if Phase 2 passes, only if net energy has been measured and independently confirmed, do we build a demonstrator engineered for continuous operation. By this stage the scientific risk is largely retired and the work becomes engineering, which is exactly when private and strategic capital becomes available and appropriate, so the public role shifts from sole funder to bridge and co-investor.

Three phases, two gates, ten years, on the order of €150 million in total if the programme runs the whole distance, which on the probabilities it most likely will not, because most of the time an early gate closes and spending stops. That asymmetry, modest committed cost up front, large costs only unlocked by success, is the financial signature of a well-designed moonshot, and exactly what NADI's milestone model is built to produce.

What NADI is actually being asked to put at risk

Separate the headline number from the number genuinely at risk. The full ten-year figure, around €150 million, is the cost only in the world where every gate is passed and the demonstrator is built, which is the world where the programme has succeeded and private capital is queuing to share Phase 3. The amount needed to find out whether the science is real, the at-risk research money, is the Phases 1 and 2 total of about €80 million over six years, and even that is shared. With Nordic partners and European research funding carrying part of the basic-science load, the realistic Dutch public commitment through NADI is on the order of €50 million over six years, released in tranches, each earned at a gate.

It is worth putting €50 million on the right plate, because in isolation it sounds like a great deal of public money and in context it is not. It is roughly what the Netherlands already forgoes through a single tax break, the innovatiebox, every six days. It is a small fraction of the €1.1 billion the country committed to PhotonDelta, its integrated-photonics programme. And it is a fraction of the €500 million already reserved for NADI in the coalition agreement — an amount NADI's advocates argue should ultimately grow to €1 to 2 billion over five years, across roughly twenty programmes. A €50 million research commitment, staged over six years and released only against results, is a single programme of ordinary size for such an agency, distinguished not by its cost but by the magnitude of what sits on the other side of the gates.

Why the Netherlands can actually build this, and why it is the credible referee

A plan is only realistic if the country proposing it can execute it. Here the Netherlands is in a genuinely unusual position, and I want to be precise about where our strength is real and where the common boast is hollow.

The hollow version is "we have ASML, therefore we lead the world in nanofabrication." We do not. ASML builds the machines; the mass-production know-how that turns those machines into chips lives in Taiwan and Korea. If this programme ever reached true mass manufacturing, that stage might happen elsewhere, and we should be honest about that from day one.

But mass manufacturing is not what frontier physics needs. It needs world-class research fabrication and, above all, world-class measurement, the ability to make exquisite one-off nanoscale structures and to measure their energy budget so precisely that a claimed effect can be believed or killed. That we have in depth. The Holst Centre in Eindhoven is a permanent joint venture between the Dutch research organisation TNO and the Belgian nanoelectronics institute IMEC, Europe's leading semiconductor research institute, sitting on Dutch soil, which means IMEC is in effect already here. QuTech in Delft holds some of the world's foremost expertise in superconducting quantum circuits, the very technology behind the Chalmers result. AMOLF's nanophotonics, the shared cleanrooms of the High Tech Campus, the PhotonDelta cluster, and the Casimir-named institute at TU Eindhoven complete a set of capabilities that exist, are among the best in Europe, and sit within an hour of one another.

This is the deeper point about Casimir Inc. and everyone like it. The defining weakness of the entire field is not a shortage of bold claims; it is a shortage of trust. Every result is a self-reported trickle that no independent laboratory has audited. What the world lacks, and what this ecosystem is built to provide, is a rigorous, independent, openly published referee with the metrology to settle the underlying question rather than merely assert an answer to it. In a field whose central problem is that nobody believes the measurements, the credible referee is worth more than the first mover. The Netherlands is positioned to be that referee. Almost no one else is.

It is worth being precise about what "referee" means here, because it resolves a tension a careful reader will otherwise notice between this open role and the security posture I set out below. There are two separable questions. The first is whether the effect is real — whether the vacuum can be made to yield net energy at all. That question is settled with published physics, replicated openly, and answering it credibly for the world is the referee's job. The second is how to build a good device — the specific geometries, materials, and efficiencies that turn a real effect into useful power. That is the part worth holding closely if the answer to the first question is yes. It is the same division that has governed nuclear technology for eighty years: the physics that fission works is in every undergraduate textbook, while the engineering of an actual weapon is among the most tightly held knowledge on Earth. Being the open referee on "is it real" and the guarded owner of "how to build it" are not in conflict. They are two layers of the same programme.

And we have the rarer thing, proven coordination. The Brainport region's habit of aligning government, industry, and academia, forged out of the near-collapse of Philips in the 1990s, has already delivered the €1.1 billion PhotonDelta programme and the €615 million Quantum Delta programme. The muscle to run a coordinated, multi-institution, hundred-million-euro technology programme is not a thing we would have to invent.

Why we should not do it alone: the Nordic partnership

The single most important piece of expertise this programme needs, the ability to make and measure the dynamic Casimir effect, lives at Chalmers in Sweden, built over fifteen years by the group that did it first. That cannot be bought or rebuilt quickly; it can only be partnered with. Sweden also brings a deep tradition of patient, long-horizon technology capital, in its great research foundations and venture firms, exactly the kind of money a demonstrator phase would need. The division of labour writes itself: the Netherlands brings fabrication, coordination, QuTech's superconducting strength, and the historical claim; Sweden brings the people who first made light from nothing and the capital culture to carry a success forward. A Dutch-Nordic programme also reads internationally as a serious regional initiative rather than one small country's eccentric gamble, which matters for talent, for European co-funding, and for surviving across electoral cycles. Sweden is, since 2024, a NATO member, which matters for the security regime below. Denmark, Finland, and Norway are natural later additions.

How industry contributes, and what it gets in return

Here is the part that most cleanly separates a real plan from a proefballon, because it is the part that usually goes unsaid. The early phases need scarce things universities do not have in sufficient quantity: industrial-grade cleanroom time, specialised metrology, and the hands of engineers who actually know how to fabricate and measure at this level, the people who work at ASML and across the Brainport supplier cluster. They do not arrive because the cause is noble. They arrive when the contribution buys a defined stake in the upside. So the programme is built, from the start, to convert contributed expertise and equipment into ownership.

The instrument is a dedicated programme entity, a purpose-built company that holds the foreground intellectual property the programme generates. NADI funds and governs; the entity owns what it creates; and that ownership is what partners, the state, and the Swedish side can hold stakes in. This is not a departure from the model. DARPA's own performers are routinely purpose-built companies, and the flexible contracting NADI is copying exists precisely to fund entities like this rather than force everything through a university or a ministry. A programme entity is what makes the bet ownable, and ownability is what unlocks the country's best industrial talent, who will join a venture with a cap table but not a civil-service grant.

Contributed equipment, cleanroom time, and engineer-hours are valued and converted into two forms of return. The first is equity in a real applications business, the advances in Casimir-force control that matter directly for microelectromechanical systems, sensing, metrology, and quantum hardware, the domains these companies already sell into. The second is a licence to the intellectual property in the contributor's own field of use within that business. Crucially, the strategic core, energy generation and defence, is treated differently: it is carved out to the nation, not shared out to private contributors, and if it ever materialises the state takes it at independent fair value. Contributors get a genuine business and their own commercial domain; the civilisation-scale prize, if it comes, belongs to the country that paid to chase it. Each partner gets the thing it actually wants, and the crown jewels stay national.

There is a deeper reason the right companies should want in, and it is self-interest of the most serious kind. Europe's competitiveness, and the long-term survival of its most strategic firms, increasingly depends on abundant, cheap, sovereign energy, both for ordinary industrial reasons and for the enormous and growing electricity appetite of artificial intelligence. A company like ASML sits at the heart of a European technology base whose future is bounded by whether this continent can power itself competitively. For such a firm, a stake in even a low-probability path to abundant domestic energy is not charity. It is a hedge against the single dependency most likely to constrain its own future. The pitch to industry is not "help us." It is "this is insurance on a risk you already carry, and we are offering you a stake in it."

National security: defence-grade from day one

The involvement of a defence agency on the American side is a signal, and here is what it signals. If this question is real, the answer is not an ordinary scientific result. It is the most strategically decisive technology since the splitting of the atom, and it cannot be run as open, publish-everything science without simultaneously handing it to every rival the moment it shows promise. The United States understood this instinctively; its vacuum work lived inside its defence research agency. The Netherlands should follow that logic from the first day, not bolt it on later in a panic.

That does not mean secrecy swallowing everything, which would be impossible and self-defeating, since the fundamental physics is textbook material and the programme needs open scientific talent. It means two layers. The outer layer, the fundamental physics of the Casimir effect and vacuum fluctuations, stays open, published, and collaborative. The inner core, the specific geometries, fabrication methods, efficiency data, and above all any result bearing on net energy extraction, runs as a classified, security-cleared programme from the outset, within a NATO and EU trusted-partner perimeter, which is precisely why a NATO partner like Sweden is the right collaborator and a non-aligned one the wrong one. This is the same principle by which nuclear physics is taught freely while weapons design is among the most closely held knowledge on Earth. NADI is, in any case, being built with a defence dimension, and the broader push around it envisages a reservation of part of the defence budget for breakthrough defence technology. A defence-grade vacuum programme is native to the agency's design.

This posture also answers the most cynical objection, "why not let the Americans do it and buy the result." If it works, it will not be for sale, any more than a warhead design is for sale. The only way to hold the advantage is to own the programme.

Who would do the work, and what this is not

The talent is identifiable and reachable: the Dutch nanophotonics and quantum-device groups at AMOLF, QuTech, and TU Eindhoven; the Chalmers group in Sweden; and, as advisors, the small international community that has driven Casimir-force control and vacuum-energy work over the past fifteen years. Names and institutions are in Appendix C.

Three disclaimers, because a serious plan states its own boundaries. This is not a perpetual-motion scheme; we are testing whether a measured, manipulable form of energy can be accessed, with explicit criteria for declaring that it cannot. This is not pseudoscience dressed as research; the fundamental layer is open, peer-reviewed, and replicated, and the whole design is built around independent verification of exactly the kind the current claims lack. And this is not a guaranteed success; the probabilities are stated honestly, the most likely single outcome is a well-documented "no," and the entire design exists to make that "no" cheap and useful rather than ruinous.

Part V: The cost, and the prize

The cabinet's reasonable question is not "is this interesting?" but "is this a responsible use of public money?" To answer it I have to be more detailed than a manifesto usually is, about both what we would spend and what we might get.

Where the money goes, and what flows back

The roughly €150 million full-distance figure decomposes into the ordinary costs of serious experimental physics, and the great majority of it is salary for highly trained people working in the Netherlands. The detailed tables are in Appendix B. The shape: Phase 1 (~€30M) is mostly people, around forty researchers and technicians, plus cleanroom time and fabrication; Phase 2 (~€50M) expands the team and pours more into fabrication cycles, precision calorimetry, independent replication, and an early patent portfolio; Phase 3 (~€70M) is dominated by demonstrator engineering and is the phase most likely to draw private co-investment, so the public share of it is the smallest part of the public exposure even though it is the largest gross number. Money is released by gate, not by calendar; if a gate does not open, the later budgets are never spent.

A research programme spent overwhelmingly on Dutch salaries is not money that vanishes. It recirculates, and a large fraction returns to the state through income tax, employer social contributions, the VAT on what well-paid staff spend, and the onward tax paid by suppliers and the people they employ. A conservative accounting of the roughly €80 million of at-risk research money recovers a substantial share, plausibly more than half, before a single scientific result is valued, and before counting the trained people, the instruments, and the intellectual property as assets rather than costs. The real cost is a fraction of the apparent cost, and the apparent cost is already modest.

The prize, measured against NADI's own bar, with solar as the witness

NADI asks each programme to return something like seven to ten times its cost in societal spillover. Most candidate programmes strain to argue they will clear that bar. This one clears it in the cases where the physics fails, and in the case where it succeeds the multiple has many more digits.

Take the failure case first, the most likely case, and recall the second lesson of solar: the revolution can live in the applications rather than the original effect. Even if the vacuum yields no net energy, the programme produces a body of capability in Casimir-force control, and that control is becoming a central problem of the next generation of nanomechanical devices, where the same vacuum force this programme studies is the thing that makes the tiniest machines stick together and fail. The ability to tune and defeat it has direct application in the microelectromechanical systems inside every phone and car, in next-generation sensors, in metrology, and in quantum hardware. Just as the photovoltaic effect's value arrived through its scaled application rather than through new fundamental physics, the value here may arrive through Casimir engineering regardless of whether the grand question returns yes or no.

I should be honest about that comparison rather than hide behind it, because it cuts both ways: solar's payoff took decades and staggering global manufacturing investment to arrive, and I am not promising the failure case returns seven to ten times its cost on a short horizon. What I am claiming is narrower and, I think, safe. Between the trained specialists released into Dutch industry, the permanent uplift in national fabrication and metrology capability, the patents in Casimir-force control, the strengthened Nordic alliance, and a definitive published answer to a question the world wants settled, the failure case returns a real and substantial multiple of a modest outlay, not zero and not a rounding error. The floor on this bet is unusually high. Whether it reaches NADI's full spillover bar depends on those applications maturing or on the science succeeding outright; the point is that you are not choosing between a jackpot and a total loss. You are choosing between a jackpot and a solid consolation prize.

Now the success case, the few-percent world. Global spending on energy runs to several trillion euros every year, and energy is the input cost underneath every other industry on Earth. A genuine ability to draw usable power from the vacuum would not capture a slice of that market; it would, over time, reconstitute it, with zero fuel, zero emissions, and no supply chain to be blockaded. The strategic dimension is starker still. Whoever first holds such a technology holds the licensing, the know-how, and the training pipeline everyone else must come to, the position the Netherlands already occupies in miniature through ASML's chokehold on lithography, transposed to the foundational input of the entire economy. Against a value of that magnitude, the seven-to-ten-times bar is not cleared, it is rendered almost meaningless.

The expected value follows from putting the two together. Multiply the substantial-but-bounded value of the likely failure by its high probability, add the enormous value of the unlikely success by its small probability, and the second term looms large despite the small probability, because the magnitude on the other side is so vast.

At this point a sharp reader raises the obvious objection, and it deserves a straight answer rather than a dodge. "A tiny probability times a gigantic payoff," they say, "can be used to justify absolutely anything. Promise me a large enough prize and I can make any lottery ticket look rational. This is just a mugging dressed up as arithmetic." The objection is correct as a general warning and wrong as applied here, for two specific reasons. First, the probability is not a number I invented to win an argument; it is anchored in physics that has been measured for seventy years and manipulated in the laboratory in the last fifteen, by named groups, in peer-reviewed journals. This is not a one-in-a-quadrillion fantasy conjured to overwhelm your judgement; it is a few-percent chance grounded in real, moving science, which is a completely different kind of object. Second, and more importantly, the mugging only works when the downside is ruinous, when you must hand over everything for a lottery ticket that probably burns. Here the downside is capped and largely recovered: the failure case returns trained people, instruments, patents, and an answer. You are not being mugged when the worst outcome still leaves you meaningfully better off. Strip away either feature — make the probability imaginary, or make the loss total — and I would drop the argument myself. It survives precisely because neither is true.

So you do not need to believe the long shot will land. You need only accept that the cost of asking is small and largely recoverable, that the probability is genuinely non-zero rather than literally impossible, and that the prize is enormous. The stories at the start of this essay are what establish the middle point: "impossible," when it rests on a shaky assumption about a poorly understood phenomenon, is not the same as zero.

The cost of not asking

One more entry on the ledger, the one easiest to leave out because it never shows up as a line item: the cost of letting someone else answer first. The question will be answered eventually, by someone. DARPA has funded it; a US startup is now built around it. The only thing in doubt is whether the Netherlands is in the room, holding the knowledge and the patents and the trained people, or whether we read about it afterwards and pay licence fees to whoever got there first. We did not turn the photovoltaic effect into the cheapest electricity in history by being the most cautious nation in the room; we watched much of that value be captured elsewhere, and we are still writing reports about why Dutch science so often fails to become Dutch industry. NADI exists, in the government's own framing, precisely to close that gap. A programme where the discovery would be the most valuable in a century, and where we hold heritage, fabrication, and metrology advantages few others can match, is the sharpest possible test of whether we mean it.

Part VI: The ask

NADI's first programmes are chosen this year, while the agency is still deciding what kind of agency it is going to be. That is the window, and it does not stay open. So let me put the ask in the present tense, as concrete steps rather than aspirations.

In the founding programme round, an Innovation Director is appointed with a Phase 1 mandate and a €30 million gated envelope: a Dutch-led, Nordic-partnered effort, run on NADI's own model of competing teams and tranche funding released only at hard go/no-go gates. The Phase 2 money is committed now but locked, released only when an independently replicated Phase 1 result opens the gate, so that today's cabinet binds tomorrow's to evidence rather than to hope, and no successor has to find the courage to continue or the courage to stop. If Phase 1 fails, the programme closes and its people and equipment redirect to Casimir-force engineering, where the capability pays for itself. The decision does not wait for certainty, because certainty is the one thing the programme exists to produce.

Around that, four things.

That the agency commit, in the first instance, only to the research phases, on the order of €50 million in Dutch public money over six years, matched by Nordic partners and European research funding, with the larger demonstrator phase explicitly conditional on the science being proven first and expected to draw private and strategic capital when it is.

That the programme be structured from day one around a dedicated programme entity that holds the commercial IP, so that the Netherlands' unique industrial assets, ASML and the Brainport supplier cluster, can contribute equipment and engineers in exchange for equity and field-of-use IP rights, while the strategic core, energy and defence, is carved out to the nation and the state's control rests on narrow, purpose-built national-security instruments rather than a broad veto, with NADI's research money kept as a grant and any returnable public investment made separately through InvestNL.

That it run as a defence-grade national-security programme from the outset, two-layered: open fundamental physics on the outside, a classified, security-cleared core within a NATO and EU trusted-partner perimeter.

And that the government use the convening power NADI is meant to wield to align the relevant Dutch institutions, TNO and Holst, QuTech and Delft, AMOLF, TU Eindhoven, around a single programme rather than a dozen disconnected grants, and to open a government-to-government conversation with Sweden about joining as a founding partner.

What the agency would not be asked for is worth stating too. Not a promise that it will work; the probabilities are on the table. Not an open-ended cheque; the gates are the opposite. Not a new bureaucracy; the coordination machinery exists in Brainport and the operating model exists in NADI's own design. The programme is built to be cheap to stop and valuable even when stopped.

An invitation

This essay is the public case. Behind it sits a confidential project plan: the ownership and capital structure of the programme entity, the mechanism for valuing and converting industrial contributions into equity and field-of-use licences, the classification ladder and security architecture, and the governance and national-security control terms, with an indicative term sheet for founding partners and co-investors. Its author holds no stake in the venture and seeks no role in it; this is advocacy, and the person making the case has arranged to be among those who do not profit from the answer.

If you are reading this from within the Dutch government, NADI, or a relevant ministry; if you are a senior figure at ASML or one of the companies in its orbit; or if you are an investor with the patience and risk appetite this requires, I would welcome a direct and confidential conversation. The detailed plan is prepared and ready to share with those in a position to act on it.

The four researchers who broke the four impossibilities in this essay had one thing in common. Somebody let them try. The Netherlands is now deciding whether to build an agency whose entire reason for existing is to be the somebody that lets people try, and its first choices will tell us whether that was a real decision or a press release. Here is a question worth its founding: measured in our own laboratories, predicted by our own Casimir, defended by exactly the kind of assumption history keeps overturning, chased right now by others with more noise than rigour, and still waiting for the one country with the heritage, the tools, and the credibility to settle it properly.

The Netherlands does not have to do this. The question will be answered eventually, by someone, somewhere, and the country can read about it later like everyone else, and pay for the result like everyone else. That is a legitimate choice, and it is entirely the cabinet's to make. Nobody should be pushed into a moonshot.

But we are the country that measured the energy in empty space. The laboratories are here. The heritage is here, the fabrication is here, the credibility to settle a contested question honestly is here. The agency is all but here — reserved, designed, a decision away. The only thing not yet here is the decision itself.

We are late. We are not, yet, too late.

Appendices: the evidence

The body makes claims. These appendices source, cost, and lay them out for anyone who wants to check the work. A realistic plan should show its arithmetic.

Appendix A: The programme at a glance, with decision gates

Funding is released by gate, not by calendar. If a gate does not open, the later budgets are never spent.

Appendix B: Cost structure and fiscal flow-back

Phase 1 (years 1–3), ~€30M: personnel ~€12M (~40 FTE); equipment & fabrication ~€10M; infrastructure & Chalmers partnership ~€4M; operations & travel ~€2M; contingency ~€2M.

Phase 2 (years 3–6), ~€50M: personnel ~€18M; fabrication & iteration ~€20M; precision measurement & replication ~€6M; IP & patents ~€2M; operations ~€2M; contingency ~€2M.

Phase 3 (years 6–10), ~€70M: demonstrator engineering, testing, commercialisation preparation; expected to draw private and strategic co-investment, reducing the public share. Funded only if Phase 2's gate is passed.

Fiscal flow-back (research phases). The ~€80M of Phases 1–2 is spent overwhelmingly on salaries paid and taxed in the Netherlands, on Dutch suppliers, and on local activity. A conservative accounting of income tax, employer social contributions, VAT on staff spending, and onward supplier tax recovers a substantial share, plausibly more than half, before any scientific result is valued. Because money is gated, the expected public spend is far below the full-distance figure: in most of the probability mass an early gate closes and spending halts.

Appendix C: People and institutions

Dutch ecosystem (execution): Holst Centre (TNO/IMEC), Eindhoven, primary advanced fabrication partner; QuTech, Delft, superconducting-circuit expertise; AMOLF, Amsterdam, nanophotonics; TU Eindhoven (Casimir Institute), photonics and talent; High Tech Campus / PhotonDelta, cleanrooms and photonic integration; TNO, applied-research and government bridge.

Swedish partner (core expertise): Chalmers University, Gothenburg, originators of the dynamic Casimir effect demonstration (2011), continuing leaders in superconducting circuits; KTH as secondary academic partner; Swedish research foundations and deep-tech venture firms as the Phase 3 capital bridge.

International advisory pool: the community behind the first repulsive-force measurement, on-demand force control, and the sustained net-extraction effort, natural members of a scientific advisory board and source of visiting collaborators.

Appendix D: The recent experimental record (why the field is live, not speculative)

In fifteen years the field moved from measuring the vacuum to commanding it, and is now being commercialised on self-reported results. No one has yet built the rigorous, independent, publicly accountable programme to settle whether the command extends to net energy.

Appendix E: How this programme maps onto NADI's own criteria

Appendix F: Ownership, contribution, and security architecture (summary)

The full mechanics, valuations, and terms are in the confidential plan; this is the public summary.

Ownership. NADI funds and governs the programme as a research grant and takes no equity; in return the state receives national-security control rights and the carve-out of the strategic (energy and defence) IP. A separate public investor (InvestNL) may take returnable equity alongside private capital. A dedicated programme entity holds the commercial IP and is the vehicle in which partners and investors hold stakes. The state's strategic control rests on layered, purpose-built instruments (a narrow special share, the Dutch investment-screening regime, export control, and above all the IP carve-out), held independently of any shareholding so they survive dilution, rather than on a single broad "golden share."

Industrial contribution. Contributed cleanroom time, equipment, and engineer-hours are valued and converted into equity in the programme entity plus field-of-use licences to IP in the contributor's own domain (for example MEMS, sensing, metrology, quantum hardware). Energy and security applications are carved out and retained by the state.

Why industry should want in. Beyond direct IP and equity returns, abundant sovereign energy is becoming a precondition for European industrial competitiveness and for the electricity demands of AI. For strategic firms, a stake in even a low-probability path to it is a hedge against their own most binding future constraint.

Security. Two layers. Open: fundamental Casimir and vacuum physics, published and collaborative. Classified: specific geometries, fabrication methods, efficiency data, and any net-extraction result, run as a security-cleared programme within a NATO/EU trusted-partner perimeter, with vetted personnel, controlled facilities, export controls, and the change-of-control veto above. Modelled on the open-physics / classified-application division used for nuclear technology.

Appendix G: Principal sources

• Coalition agreement "Aan de slag", 30 January 2026 (D66–VVD–CDA minority government; €500M capital injection reserved for the "still to be established" NADI, per the budgettaire bijlage); Wennink report (2025); Ministry of Economic Affairs Kamerbrief and Ontwerpvoorstel on NADI (9 February 2026), an exploratory design proposal whose adoption remains a decision for the cabinet — source for NADI's operating model, milestone funding, portfolio approach, the 7–10x spillover target, per-programme budget band, and the four proposed themes (Digitalisation & AI; Security & Resilience; Energy & Climate technology; Life sciences & Biotechnology). The larger €1–2bn-over-five-years figure and the defence-budget reservation are the advocates' proposal (open letter, June 2026), not committed budget. On the capital-injection/Eurostat funding tension and the ~€150M/year real-budget gap: Instituut voor Publieke Economie (June 2026). Innovatiebox ≈ €3bn/year: same source.
• Casimir (1948); Sparnaay (1958); Lamoreaux, Phys. Rev. Lett. 78 (1997) — prediction, first measurement, precise confirmation.
• Munday, Capasso & Parsegian, Nature 457 (2009) — repulsive Casimir–Lifshitz force.
• Wilson, Johansson, … Delsing, Nature (2011) — dynamic Casimir effect.
• Zhang et al. (USTC), Nature Physics 20 (2024) — magnetic tuning of the Casimir force; and the 2025 comment re-examining it.
• Spreng, Shelden, Gong & Munday, "Casimir repulsion with biased semiconductors," Optica Quantum (2024) — a proposed (theoretical) method for electrical control of the force; and Shelden, Spreng, Garrett, Rahman, Kim & Munday, "Casimir Force Control Enabled by 3D Nanostructures," Nano Letters 25 (2025) 9254 — experimental control via engineered nanostructures. Both UC Davis (Munday group).
• G. Moddel et al., Atoms 7 (2019) 51 — an assessment of vacuum-extraction approaches, with later experimental work (optical-cavity-induced current) and patents; tantalising but inconclusive.
• DARPA QUEST and ARRIVE programme descriptions — vacuum engineering and energy transduction.
• On Casimir Inc.: launch reporting in The Quantum Insider, The Debrief, and others (May 2026), and the $12M oversubscribed seed round led by Scout Ventures. The underlying paper is H. White et al., "Emergent quantization from a dynamic vacuum," Phys. Rev. Research 8, 013264 (2026), DOI 10.1103/l8y7-r3rm — a theoretical result, not an independently verified device. On the founder's contested record, coverage of the NASA Eagleworks EmDrive results and the 2021 TU Dresden null result.
• IEA, World Energy Outlook (2020) — solar as, under favourable conditions, the cheapest electricity in history.
• The ~10^120 cosmological-constant ("vacuum catastrophe") discrepancy is standard in the quantum-field-theory and cosmology literature.

Programme cost, phasing, and fiscal flow-back figures are the author's indicative estimates, built from standard Dutch research-salary and cleanroom-cost assumptions, intended to convey scale and structure, not to substitute for a formal budget. Probability estimates are the author's own subjective judgements.