Sovereignty policy should focus on builders, not buyers.

Everyone treats the sovereignty debate as a demand problem: who should buy what. I don't think it should be the priority. Real sovereignty is built on the supply side, with the builders.

Over the last few months I've had the opportunity to be in many conversations with people thinking about French industrial policy and technological sovereignty for a living. I was struck by how differently people weigh the urgency and the priorities. This post is where I've netted out.

Most of what I heard is on "demand". Who should buy what, which budgets should flow where, what we should oblige the private sector to purchase. Yes, the state should buy French but it doesn't move the needle the way other levers can.

The harder problem, the one that really matters, is making the tech worth buying in the first place.

Economic patriotism can't be mandated in the private sector

In cloud, software and AI, a handful of players (mostly American) capture a massive share of the growth and it creates deep dependencies. For France that's structurally bad, sometimes dangerous. I'm not debating the dependency, I think we all agree on the diagnosis, we disagree on the answers.

These dependencies are real but we won't change market dynamics by asking private company CTOs to be more patriotic.

A company picks technology when it lowers cost or risk, or because it gives access to a better product and a better ecosystem of partners. Not because of sovereignty or national preference. Such technology gets chosen when it's mandatory (or close to), in sectors like health, defence and industry, or when it's simply better. That's why demand isn't the main lever. Buyers buy software for a very simple reason: having tools that help them solve their business problems. If a better option is a foreign option, for the overwhelming majority of buyers, it isn't a concern.

If the offering isn't good enough, we should diagnose why and treat the causes

You're sovereign when you can build and commercialise a best-in-class offering. Full stop. The rest follows from that. French preference should be downstream from French excellence. When French or European technology isn't at that level, we have to diagnose why.

The gap isn't our engineers. They're everywhere in the leading US companies so it isn't a competence problem.

The problem is that too many European founders have not had the means to recruit the very best people, pay them at globally competitive rates and keep them long enough to turn promising companies into global leaders. Leaders ambitious enough that joining them feels as exciting as joining companies like Nvidia.

In cloud or infrastructure software, the gap was built over ten or fifteen years and scale plays a big role. When you join AWS, Netflix or Google, you work on a set of problems (volumes, distributed systems) that you otherwise wouldn't have access to in Tier-2 or Tier-3 competitors. As a young engineer, it's a question of learning environment and being exposed to what great looks like (btw: we happily back those profiles and some of them do end up starting here in Europe).

Generative AI is a different story in my mind. The market really opened up just 3-4 years ago. OpenAI, Anthropic, Mistral and the others are attacking broadly comparable problems. The gap isn't a historical scale gap, it comes very directly from funding, access to compute and the ability to pay the best people at globally competitive rates.

That's exactly where the state needs to be far more pro-active.

You can talk about sovereignty all you want but the single biggest thing hollowing it out is brain drain. If the brightest minds of our country can work on more ambitious problems and/or earn far more elsewhere, we can't act surprised that part of the value gets built... elsewhere. Trump turned out to be the best talent-repatriation argument we've ever had and it's a little sad that we owe it to that.

Talent is the clearest element this whole sovereignty debate keeps missing. No amount of demand policy or mandated public contracts will keep your best people from leaving to San Francisco.

When writing this post, that's the split I keep coming back to:

• Demand-side sovereignty asks who should buy the technology and how politics can affect those decisions
• Supply-side sovereignty asks whether we can build tech well enough in the first place so that buying it is the obvious decision

The first is talking about a purchasing policy. The second is setting the conditions of successful production: the capital, the compute, the talent, the ability to build a company here as ambitious as its global rivals.

I'm convinced our policy efforts spend too much energy on the first while most of the difference is made by the second.

The state should create the conditions, not pick the stack

The state already has plenty of levers.

• It steers public procurement.
• It funds R&D.
• It supports its champions through diplomacy and soft power.
• It creates anchor funding and demand in critical sectors (France 2030 did exactly that on low-TRL bets like quantum or deep-sea exploration).

It already does all of this and does it well on many dimensions.

For the interventionists here, there are many ways to go further to push sovereignty. Conditioning subsidies or public contracts on the use of sovereign building blocks could be an idea? In my view though, it's a slippery slope. Asking the state to pick the technical stack of private companies and having the state as pseudo-CTO? No bueno.

In critical sectors, you can take a more directive stance. There the state is legitimate in requiring that part of the stack be sovereign. For the general private sector, if the home-grown alternative is worse or less mature, we shouldn't be surprised that it isn't chosen and we should make sure we understand why.

The only way a truly sovereign tech sector can emerge is by helping the best French and European players have access to all the ingredients to global success that I mentioned above.

There are three layers we tend to blur together.

• The first is fair competition. Do some practices by dominant players artificially block alternatives from emerging? If so, go after them with the international legislative arsenal.
• The second is competitiveness. Is the French or European competitive with global alternatives? Is it better, equivalent or simply good enough for a given workload?
• The third is industrial policy. If the national/sovereign offering isn't globally competitive yet, what needs to be done to get it there?

What the state could actually do

If we genuinely want to strengthen digital sovereignty, the third debate is the one to act on, and it has four fronts.

Talent
If we want to build global leaders, we have to be able to retain and attract the best global talent.

• Lower the all-in cost of hiring critical technical profiles. You can focus it on critical sectors: AI, quantum, cybersecurity, semiconductors, defence, robotics, compute, infrastructure software. If being competitive in those fields is truly a national priority, ditch the employer contributions or cap them at a low level so that companies can hire more competitively.
• Keep the Credit Impot Recherche and CIFRE mechanism but stop using it as a crutch that only helps us avoid being totally relegated vis-a-vis global standards. Speak to any global founder that has opened or considered opening an office for research in France to hear their initial thoughts until they learn about these incentives.
• Fast talent visas. We're already doing great with the French Tech Visa but we should enhance it with a competitive tax regime to continue attracting the best talent.

Capital and compute
In AI, cloud, semiconductors or infra, good engineers aren't enough, you need massive amounts of capital. These categories need far more capital than traditional SaaS. Access to competitive compute is a key part of the equation. If our players fight with less capital, less compute and weaker comp packages, they will naturally struggle.

A more contestable market
Rather than imposing vendors, the state should focus on keeping healthy market rules: data portability, interoperability, open standards, advocate for more open source. When bundles, credits or pricing clauses artificially make it tougher to compete, act, but with precision. Not every bundle is anticompetitive just as not every scale effect is illegitimate.

Public demand
This is where buying French belongs and it's deliberately last.

• We have to be ok with digital sovereignty as an industrial doctrine. If we want 5/10/20% of some budgets to flow to European players, it has to be said clearly. These sovereign obligations should only apply to the sectors that justify them otherwise it's just more bureaucratic friction.
• Performance requirements should also be tied to it to avoid regulatory capture from inferior products.
• Public procurement should be a genuinely demanding first customer rather than a shelter from competition.

What the state can't do: speedrunning cultural change

Corporate buyers in the US buy software from emerging companies faster and with more risk tolerance than their EU counterparts.

The gap is culture (not engineering or product). More than 30 years of software purchasing normalised the idea in the US that a three-person startup could be a legitimate vendor. No amount of government purchasing mandate will change this.

The US is still where the majority of enterprise software spend happens. The best French AI companies will likely be transatlantic before they're European if that's where their fastest and most sophisticated buyers are based. "Buy French" is at best a narrow answer. We should focus on making European companies the obvious choice thanks to their excellence.

YES to creating the conditions for champions to emerge. Capital, compute, talent, fair competition and demanding first customers are all necessary.
NO to protected mediocrity.

Real sovereignty is local companies winning on merit.
At home and abroad.