After SCAF: Who Will Pay for the Future Rafale?

On September 22, 2026, Dassault Aviation announced that it had flown two “sovereign” artificial intelligence algorithms aboard a Rafale: one designed by its own engineers, the other developed with Thales and its cortAIx center. The press release runs four paragraphs and says almost nothing. But read alongside what the DGA, France’s defense procurement agency, confirmed ten days earlier, it tells a remarkable story: France has taken back control of its combat aviation. Alone.

I am genuinely glad. And that is exactly why I refuse to stop at applause.

What I Wrote in February, and What Has Changed

In February, I argued that Dassault was defending France better than the Élysée by blocking SCAF, the Franco-German-Spanish fighter program known in English as FCAS, and that a Plan B existed, but came with a price tag. We are there now. The Franco-German negotiations have collapsed, and the Ministry of the Armed Forces is holding its course: the money set aside for the future fighter will go to a national successor to the Rafale. Éric Trappier had already said as much before the Senate’s defense committee on July 1, 2026: “Plan B is a super-Rafale.” A transitional Rafale F5, backed by a combat drone, which in his view would outperform the F-35 at a lower cost.

Meanwhile, the building blocks are falling into place. Five contracts have been awarded to prepare the Rafale F5, expected in 2033 at the earliest: a new radar, a jam-resistant data link, a fully digital electronic warfare suite, and the preliminary design of the M88 T-REX engine, 20% more powerful than the current one within the same envelope. In July, a Rafale F4 carried out its first collaborative combat trial with an electronic warfare drone developed alongside the French startup Harmattan AI. And now, AI has entered the cockpit.

The prime contractors are French, and things are moving. Not everything underneath is French yet: some of the chips, design tools, and electronics still come from elsewhere, much like the Nvidia processors on which even our AI champion depends. That is precisely what makes the rest so valuable. And one thing is still missing.

A Plane That Thinks, Provided It Can Be Fed

The only technical sentence in Dassault’s press release is also the most interesting: mastering these algorithms requires resource frugality on an embedded platform operating under severe constraints. Translation: there will be no ChatGPT in the Rafale, and that is excellent news.

A fighter jet in contested airspace cannot rely on the cloud. It is no accident that the F5 is getting a data link built to resist jamming: wherever it goes, communications will be degraded, if not cut entirely. Any intelligence useful to the pilot must therefore run on board, powered by the electricity the engines generate and cooled by whatever the airframe allows. Every watt of computing is paid for in thrust, in range, in infrared signature. The Americans know this well: thermal and power management is one of the F-35’s most expensive headaches.

This is the one arena in AI where American financial firepower is not enough. Gigawatt-scale data centers do not fly. In a cockpit, the question is no longer “who has the biggest model?” but “who gets the most intelligence out of every watt?” I have argued elsewhere that the “good enough, fifty times cheaper” principle is grinding down the economics of American AI, and that Apple has triumphed in on-device AI at the cost of deliberate trade-offs. The military, for its part, never believed in gigantism.

Doing more with less is the story of French aviation. The Rafale is compact, its M88 engines are modest next to their American counterparts, and aerodynamics and versatility make up the difference. That culture of constraint is exactly what embedded AI demands. I am not claiming France is ahead: nobody knows what these two algorithms actually do, and the Americans are flying autonomous pilots too. I am saying that on this particular field, we are finally playing on equal terms.

We still need an aircraft to carry them. And an engine to produce the watts.

The Check Nobody Has Signed

This is where enthusiasm collides with reality. The updated military programming law mentions a national sixth-generation fighter demonstrator by 2035. It does not provide the money for it.

Two engines need to be distinguished here, which the DGA mentioned only a few sentences apart. The T-REX is for the F5, and its studies have already been contracted. The billion I am talking about is for what comes next: an engine with an entirely new architecture, targeting some 11 tonnes of thrust, for the sixth-generation fighter. For that one, the DGA has put the cost of studies at one billion euros. Safran is offering to cover between 100 and 150 million of it, and the head of the DGA himself admits the negotiations with the engine maker are tough.

Meanwhile, Éric Trappier is urging the state to commit right now, so that France does not end up at the mercy of American developments and does not lose its expertise. He is right. But a presidential election is coming in 2027, public finances are exhausted, and recent history has taught us that France’s grand ambitions rarely die from a veto. They die from postponement.

I have written before that France prefers regulating its decline to building its future. We have finally walked away from SCAF. The question now is whether we have the means, or only the pride.

Why This Is Not Up for Bargaining

Before talking money, it is worth explaining why this particular expense cannot be haggled over like any other.

Deterrence cannot be subcontracted. The head of the DGA said it before the Senate: a future stealth fighter is indispensable because “we need it for nuclear deterrence.” The airborne component of our ultimate weapon rests on an aircraft. If a single critical part of that aircraft depends on another country, that country holds a say over the final guarantee of our existence.

A single foreign part is enough. In 2018, Washington blocked the export of SCALP missiles to Egypt because of one American component subject to ITAR regulations. A foreign supplier decided, in our place, to whom we could sell our own weapons. Controlling the supply chain is not chauvinism. It is the condition for remaining master of one’s exports, and therefore of one’s diplomacy.

Expertise dies quietly. The last French fighter designed from a clean sheet took to the air in 1986. The irony of history: that Rafale demonstrator flew on two American General Electric engines. Independence does not fall from the sky; it is built piece by piece, and it is lost the same way. Within ten years, the engineers who know how to design a fighter will have retired. Lost know-how cannot be bought back at any price.

This is not spending, it is an investment. The Rafale, once written off as unsellable, has been exported by the hundreds, from India to the United Arab Emirates, by way of Greece and Indonesia. It sustains hundreds of French companies and brings in foreign currency with a ripple effect few other industries can match.

A Quarter of a Percent

So where do we find the money? I am making a choice, and I present it as one: drawing on what we spend on the past in order to fund the future.

France devotes nearly 14% of its national wealth to pensions, roughly 400 billion euros a year. The billion needed to study a sixth-generation engine amounts to a quarter of one percent of a single year of pensions.

I know what this argument is worth: it is the argument of anyone who wants a billion for their cause. Why this billion rather than hospitals, civilian nuclear power, or ammunition? Not because the aircraft is nobler. Because it is irreversible. A hospital underfunded one year can be made whole the next. An engine industry that lets its generation of engineers walk away cannot rebuild itself in ten years, not even with ten times the money. And this billion is not a recurring expense: it pays for studies, once, for a product that will earn export revenue for forty years. Anyone is free to prefer other priorities. I am simply saying this is the only one we will never be able to make up for.

The pay-as-you-go pension system, whose dead end I have already described, is a contract between generations. It is legitimate to devote a tiny share of it to the generation that will pay for all the ones after.

In concrete terms: a contribution from the highest pensions, capped, earmarked for the program, and labeled as such. I am not advocating a general pension freeze. That is an accounting sleight of hand that hits everyone to plug any hole at all. A targeted, capped, and earmarked contribution is something else entirely, and I draw that distinction deliberately.

For working people, there is a second path, a voluntary one. Millions of French savers put their money into S&P 500 ETFs and, without realizing it, help finance Lockheed Martin. I have explained how your PEA, France’s tax-advantaged equity savings plan, buys America without ever owning it, and how France sells shares while China builds machines. That savings needs to be redirected, and the tool already exists. Not ordinary Treasury bonds: their proceeds disappear into the general budget, and no one would ever know they had paid for an aircraft. Not stocks either: buying Dassault or Safran shares on the market enriches the seller, not the program. The model is the green OAT, France’s green sovereign bond. Since 2017, the state has issued bonds whose proceeds are tied to identified expenditures, with an annual public report on how they are used. Nothing prevents the creation of a “sovereignty OAT” tomorrow on the same principle, with a tranche reserved for individual investors, in the spirit of the Pinay loan of 1952 or the Balladur loan of 1993. Every subscriber would know where their money goes, and would watch it fly.

What I Expect to Hear Back

The first objection I have already met head-on: every billion given to defense is taken from somewhere else. True, and that is why I call it a choice, not a self-evident truth.

The second is more formidable: why should taxpayers or savers finance research for highly profitable listed companies whose profits will flow out as dividends? With a sovereignty OAT, savers subsidize no one: they lend to the state, which repays them with interest. That leaves the public money paid to manufacturers for the studies. That money must not be a gift: the state should secure a return on export sales of the technologies it has financed. If the risk is shared, the reward should be too.

My Take

We have done the hardest part: saying no. No to an industrial carve-up that would have handed over our flight controls, no to co-managing an aircraft on which our deterrent depends. What remains is the most thankless part: paying.

The AI that just flew aboard the Rafale is a promise. The sixth-generation engine is its precondition. And between the two lies a billion euros that no one has yet found, in a country that spends four hundred billion every year on its pensions.

I am not asking France to choose between its elders and its children. I am asking it to remember that the former lived under a sky that others had paid to protect for them. Now it is their turn to pay for tomorrow’s.