Ask who pays for quantum computing in France and the usual answer is a number: a national plan of 1.8 billion1 euros, announced with some ceremony and repeated ever since. The number is real. It is also the least useful thing you can know about the money. In its audit published in July 2026, the Cour des comptes records that 597 million7 euros of that plan had been engaged by 31 December 2025, that 86 % of the funds were aimed at quantum computing, and, in a sentence most readers skip, that one of the State's largest commitments takes the form of a purchase contract rather than a grant. That last detail changes who the customer is.
This is the first piece of the quantum chapter of the CareerOn Industry Atlas. The chapter follows one industry from its money to its jobs, the way the earlier chapters followed semiconductors, energy and rail. This piece asks the question an investor, a policy reader or a candidate should ask before any other: where does the money in French quantum actually come from, on which dated documents, and what does each figure measure?
French quantum is funded by the State in two different roles that the headline numbers blur together, as a planner that announces envelopes and as a buyer that signs contracts, and the second role matters more for companies and careers than the first; the private money is real but thin, no admitted document gives French quantum sales, and a reader who adds the published figures together produces a total that no document supports.
We read eight dated documents for this piece and checked every quoted passage against the file itself before writing a word. Four are French public documents: the Cour des comptes audit of the national strategy, the parliamentary science office's report on that strategy from January 2022, the strategy's own annual report for 2022, and the programme page of the defence purchase programme PROQCIMA, read on 1 October 2026. One is European: the Commission's quantum strategy of 2 July 2025. Three are international: two OECD papers from 2025 and a joint OECD and European Patent Office map of the global quantum ecosystem from 17 December 2025. A press release that restated the defence programme was refused, because it adds nothing the programme page does not print. The audit court's own website refused our automated read; we used the identical file the national strategy portal hosts, and we say so here.
One plan, three numbers
Start with the plan, because every other figure leans on it. In January 2022 the parliamentary office for scientific and technological assessment described the national strategy as a plan of 1.8 billion euros of investment over 5 years, of which 1 billion came directly from the State. Four years later the Cour des comptes printed the same 1.8 billion, but with 1.5 billion euros of public funds inside it.
It would be easy to read that as growth: the State's share rising from 1 billion to 1.5 billion. It is not. The two documents measure different things. The parliamentary figure counts the State alone; the audit counts public funds, a wider category that includes more than one public purse. Neither is a sum of money spent. Both are envelopes: what was planned, not what was paid. The audit's own measure of what has actually moved is the engagement figure, 597 million euros by the end of 2025, and an engagement is a commitment to pay, not a payment.
One plan, read at launch and read by the auditor
| What the document prints | Measure | Publisher, date |
|---|---|---|
| €1.8 billion over 5 years, of which €1 billion directly from the State | Planned total at launch | OPECST, 20 January 2022 |
| €1.8 billion planned, of which €1.5 billion public; 86 % aimed at quantum computing | Planned total, audited | Cour des comptes, July 2026 |
| €597 million engaged at 31 December 2025 | Engagement to date | Cour des comptes, July 2026 |
Three readings of one plan, then, and none of them is the amount that reached a laboratory or a company. That is not a flaw in the documents. It is how public money is reported, and the discipline is to keep each figure under its own name. A reader who writes that France has spent 1.8 billion on quantum has made two mistakes in one sentence.
The first prediction, and why it failed
Before reading any document we wrote down three predictions and the evidence that would break them. The first was that public money is the larger source of quantum funding in France, larger than private money. We expected it to hold. It fails, and not because private money turned out to be larger. It fails because the comparison cannot be made from the record.
The strategy's annual report for 2022 is the clearest place to see why. It puts 350 million euros towards basic research, infrastructures, technology transfer and education, and then adds, in its own words, "Even during a private equity shortage, a remarkable amount of 290 M€ has been raised". It is tempting to set those two numbers side by side and call the ratio a public to private split. The basis does not allow it. The first is a public allocation over the life of a programme line; the second is the money start-ups raised from investors. One is a budget, the other a set of funding rounds. The audit adds a qualitative picture: France suffers from a lack of private funds and growth capital for its start-ups, there is only one French venture fund specialised in quantum, and the larger start-ups have raised Series B rounds of around a hundred million euros. That describes the shape of private money. It does not total it.
The international evidence confirms the limit rather than solving it. The OECD and European Patent Office map reports that "Government involvement in quantum investment has been highest in Australia and the Netherlands", with over 70 % of quantum investment in the period 2021 to 2024, and that France, Germany and the United Kingdom have "lower shares of government funding". That is a measure of who invested in quantum firms, built from deal data. It is not a measure of state programme budgets, and it prints no French share as a number. So the honest verdict is the one we record: the first prediction fails as worded, and we refuse to print a comparison no document makes.
The State as a customer
The second prediction held, and it is the most useful finding in this piece. We expected that the State is not only a funder of French quantum but a buyer of it. The defence programme PROQCIMA is the evidence. On 5 March 2024 the armament procurement agency, the DGA, notified framework agreements to five companies. The programme page states that the agreements run for a total of at least 10 years, for a maximum of 500 million euros.
Two cautions before the significance. The 500 million is a ceiling, not an order already placed and not money already paid. And the page is a living one: we read it on 1 October 2026, and its technical target has already been rewritten once. With those limits stated, the audit court explains why the form matters. The programme, it writes, takes the form of a public contract and not a grant, and so becomes revenue for the companies, with medium-term visibility.
The State signs as a customer, not only as a funder
| What the document prints | Measure | Publisher, date |
|---|---|---|
| On 5 March 2024 the DGA notified framework agreements to five companies, over at least 10 years, for at most €500 million | Procurement ceiling | SGPI, Ministry of the Armed Forces, read 1 October 2026 |
| It takes the form of a public contract, not a grant, and so becomes revenue for the companies | Form of the money | Cour des comptes, July 2026 |
Sources 2 Ministère des Armées · 1 Cour des comptes
A grant pays for research. A contract pays for a delivered machine, against a specification, on a schedule. For a start-up that difference is the difference between a laboratory and a supplier. It also explains a pattern that confuses outside observers: why French quantum companies talk about customers when the commercial market is still small. Their first customer of scale is the defence ministry, and it buys rather than subsidises.
Five measures that do not add up
The third prediction was that we would find a dated figure for French quantum sales. We did not. No admitted document prints one. What the audit prints instead is a share: the market share of French companies in quantum hardware stands at 20 %. That is a position in a market, not the size of the market, and it is the closest the record comes to commercial weight. We reframed the prediction rather than stretching it, and the revenue figure stays unprinted.
The wider record then offers a set of large numbers that are each correct and cannot be combined. The Commission writes that over the past five years the EU and the Member States have invested more than 11 billion euros in quantum technologies. It also cites a market forecast that grows from 2 to 3 billion euros7 today to 155 billion7 by 2040. The OECD's policy primer reports that the European Union leads in funding announcements with around 8.4 billion dollars, "combining national and EU-wide commitments", with France at 2.2 billion dollars, behind Germany at 3.3 billion7 and ahead of the Netherlands at 1 billion. A second OECD paper names a single French research programme at 170 million dollars.
Five measures of money, never added together
| What the document prints | Measure, scope | Publisher, date |
|---|---|---|
| €350 million of public money for research, infrastructure, transfer and training; €290 million raised by start-ups | Public budget and private raises, France | France 2030, March 2023 |
| Government share above 70 % in Australia and the Netherlands; lower in France, Germany and the United Kingdom | Share of quantum investment, by country | OECD and EPO, 17 December 2025 |
| More than €11 billion invested by the EU and Member States in five years; a market forecast to grow from €2-3 billion to €155 billion by 2040 | Public total, EU; market forecast, global | European Commission, 2 July 2025 |
Sources 4 France 2030 · 6 Organisation for Economic Co-operation and Development · 5 European Commission
Read together, those figures describe five different things: a French public budget, private funding rounds, a share of investment by investor type, a European public total, and a global market forecast. Some are in euros and some in dollars. Some are announcements, which are neither engagements nor spending. The forecast is global and is printed with its source; it is not a French market and not our estimate. Adding any two of them produces a number that exists in no document, which is the most common error in writing about this industry.
What the documents do not say
The limits of this record are part of the finding. No admitted document gives a consolidated total of private money invested in French quantum. None gives French quantum revenue. None says how much of the 597 million engaged has been paid out. The defence ceiling says nothing about what share of it will be ordered. The OECD announcement figures come from a consultancy tracker and are announcements, not commitments. And the 155 billion forecast, the figure most often quoted in pitch decks, is a projection for a global market fifteen years away.
We also checked our own catalogue against the record, with the same rule we apply to employers: a simulation is placed only where an admitted document names its technology. The audit names cat qubits, superconducting qubits designed to resist noise, among the funded French hardware routes, alongside silicon spin qubits. That places our superconducting qubit simulation. Our cryogenic electronics simulation is not placed: the record funds cooling systems and silicon spin qubits, but no admitted document names cryogenic control electronics, so we print it as unplaced rather than claim a link the evidence does not make.
What this means for your career
For a candidate, the money tells you where the work is before any job board does. Three conclusions follow from the record, each bounded by what it can prove.
First, follow the contract, not the envelope. The plan's headline figure is spread across research lines; the defence programme is a contract with five named companies over at least ten years. Engineers who can deliver against a specification, test a machine and document its performance are doing the work a buyer pays for, which is a different job from the research the grants fund. If you are choosing between offers, ask whether a team's money comes from a grant or a contract. The answer predicts your next three years.
Second, the hardware is where France holds a measured position, 20 % of the market by the audit's reading, and the funded routes are named. If you come from physics, microwave engineering or low-temperature measurement, the superconducting route is one the record names. You can rehearse it in our simulation Superconducting qubit engineer: designing and characterising a cat qubit, which asks you to work through qubit design, cryogenic measurement and a coherence budget the way a hardware team would. Our second quantum simulation, Cryogenic electronics engineer: integrated quantum control at 4 K, is real practice in cryo-CMOS and thermal budgets, but we do not claim the documents in this piece name that role; treat it as preparation for a profile the record has not yet counted.
Third, the record names work we do not yet simulate. The Commission lists "quantum software engineering, system integration, and quantum cybersecurity" among the skills the industry needs. We have no simulation for any of the three, and we record them as gaps in our own catalogue rather than stretch an existing one to fit. For a software engineer or a security specialist, that is a signal: the field is asking for profiles it does not yet train at scale.
For employers and investors, the same record carries a quieter lesson. A company whose first large customer is the State is building to a buyer's specification on a public timetable. That is a strength while the commercial market is small, and a dependency to plan around when it grows.
How we read these documents
Every figure in this piece appears in the passage of the document we cite, and every passage was checked against the live file before writing. Each figure keeps the measure its publisher gave it: a plan is a plan, an engagement an engagement, a ceiling a ceiling, a forecast a forecast. Where two documents print different numbers for what looks like the same thing, we print both with their dates and do not reconcile them. Nothing is summed across documents, nothing is converted between currencies, and nothing is averaged. The next piece in this chapter follows the same money into the laboratories and companies that receive it.