You spent twenty years systematically refusing to monetize VLC—rejecting buyout offers worth tens of millions of euros, advertising partnerships, or massive distribution bundles with other software—in the name of a simple principle: free software must remain a common good. You are now CTO of Scaleway and founder of Kyber, a company operating in a highly competitive, privately funded environment. How do you reconcile these two dimensions? Is there a contradiction between the ideal upheld by VideoLAN and the economic realities of tech entrepreneurship, or do you simply consider that the battle continues today on a different front?
I am not hostile to money; I simply believe we can have ethical business models. With VLC, I refused to make users the product. Advertising, tracking, dubious bundles, or selling the project would have destroyed trust, and thus VLC’s value. That does not mean refusing all economic activity. I created VideoLabs in 2012 to pay developers working on VLC, x264, or FFmpeg. Brilliant people must be able to make a living building useful technologies, not just by optimizing ads. I see no contradiction. A company can fund, industrialize, and sell a technology without capturing or degrading the common good it depends on. At Scaleway, we are building European infrastructure. With Kyber, we are building a product for a real market, based on open source. The structures are different, but I am always looking for the same thing: building real technology, staying independent, and making money without damaging what we are building.
In 2021, when Shadow was placed in receivership, you participated in a takeover attempt against the offer led by Octave Klaba. The court ultimately chose another path, and you left the company a few months later. Three years after this episode, you became CTO of Scaleway, a subsidiary of the Iliad group. Looking back, what do you take away from this sequence? Did this experience teach you anything about the balance of power between technological vision, funding, and economic sovereignty?
Shadow taught me that excellent technology cannot compensate for a bad business model or poor governance. When I arrived, the problem was no longer primarily technical. The hardware was too expensive, some contracts were bad, the economic model was fragile, and the teams were not being listened to enough. At that stage, product quality is no longer enough. You need capital, discipline, a good business model, and shareholders who understand what they are funding. I participated in the takeover attempt because I believed the technology, and especially the teams, deserved to be saved. I still think our offer was the best on these two points, especially when you look at the result today. Sovereignty is not decreed. If you do not master capital, governance, and the long term, you do not truly master technology. You can have designed it and still lose it.
You have often criticized "trusted cloud" initiatives like Bleu or S3ns, arguing that they do not fully meet the requirements of true digital sovereignty. Yet, these projects benefit from the support of the State and major industrial groups. Where do you draw the line today between pragmatic compromise and strategic surrender when it comes to building a European digital infrastructure?
The State does not understand what it is doing: SSH does not stop at borders. So-called "trusted cloud" offerings provide neither real sovereignty nor complete trust. An American technology operated by a French company and certified by ANSSI may improve operational security or legal protection. Very well. But it remains fundamentally American. The original provider can modify the product, its prices, its licenses, or obey an American political decision. The United States also retains legal means of access to the data. A compromise can be defended to respond to an emergency. In that case, it must be stated, treated as transitional, and accompanied by simultaneous investment in a European solution. Using public money to permanently install an American dependency is not pragmatism. It is an industrial surrender, a grave error.
You regularly explain that European technological dependence is not just about software or AI models, but also the most fundamental layers of industry, notably semiconductors and inference chips. Yet, Europe seems to have invested more in AI models than in the material infrastructures that make them possible. Are we witnessing a form of collective denial? And what, in your view, should be the industrial priorities to allow Europe to concretely reduce this dependence?
Yes. We readily fund what is visible and easy to announce: a model, a startup, a benchmark. But AI does not start with the model. It requires electricity, data centers, networks, memory, semiconductors, packaging, compilers, software, and then machines capable of training and executing the models. Europe will not catch up on ten years of delay with a new plan and a few billion distributed without coherence. It must rebuild a complete chain: chip and memory production, computer assembly, energy, data centers, CPU, GPU, NPU, and network chip design, followed by all the software needed to use them. The order of magnitude for investment is 100 billion euros, minimum. We do not need to copy Nvidia or TSMC exactly. We need to be able to continue functioning when a foreign actor decides to stop selling to us, change its conditions, or serve its own market first. Without European public and private procurement, this industry will not take off, and we will not have autonomy.
Projects like VLC, FFmpeg, x264, or dav1d have become essential building blocks of the global digital economy. Yet, these infrastructures often rely on a limited number of developers and fragile economic models. After several recent incidents in the open-source ecosystem, the question of maintenance and funding has become central. How can we guarantee the sustainability of these critical projects? In your view, who should assume this responsibility: user companies, states, foundations, or the community itself?
The primary responsibility lies with the companies that use these building blocks on a large scale and make a lot of money from them. They must fund the projects recurrently, not just pay for an audit when a vulnerability becomes visible. The community must retain technical direction. Foundations can hold assets, contracts, and neutrality, but they must not become an end in themselves. Too much governance, compliance, and meetings can kill the project they claim to protect. Critical work is rarely spectacular: reviewing patches, maintaining continuous integration, making releases, porting code to new architectures, responding to reports, documenting, and training new contributors. This is what prevents a project from dying. The State can help through public procurement and maintenance funding, but it will never replace the maintainers. And today, maintainers are under pressure.
You preside over VideoLAN, direct the technical strategy of Scaleway, develop Kyber, and continue to intervene in numerous open-source projects. Behind these successes lies a rarely addressed question: that of transmission. Who will take over VLC and the major projects you have helped build tomorrow? More broadly, is France still training enough engineers capable of developing and maintaining the critical digital infrastructures on which a large part of our economy now depends?
You do not transmit VLC by appointing a successor. VLC is not my personal project. Many developers already work on it immensely, and the project belongs to them as much as it does to me. Transmission means bringing in new contributors, giving them responsibilities, and avoiding making a single person indispensable. France still trains excellent engineers. However, it too often pushes them toward management, consulting, finance, or sales, because technical careers are less recognized and less well paid. We say technology is strategic, but salary scales and organizational charts tell the opposite story. An engineer must be able to remain an engineer their whole life, progress, make decisions, and be very well paid without having to manage fifty people. We also need to return to more practice: reading code, writing it, maintaining it for ten years, and assuming its consequences.
In your podcast "À La French," you regularly discuss the geopolitical developments reshaping the global Internet. We are today witnessing a progressive fragmentation of the digital world between American, Chinese, and European spheres of influence. In this context, how can open source preserve its universal character? Is the idea of an open and global Internet still realistic, or must we prepare for a more fragmented digital world?
Open source does not float above geopolitics. Developers live in countries, projects depend on funding, code forges belong to companies, and compilers run on infrastructures subject to national laws. But open code remains our best protection against total fragmentation, and our best chance to be independent of suppliers: a proprietary service can disappear or be cut off overnight; open code can be audited, copied, modified, and maintained elsewhere. The global Internet will likely become more fragmented. We must therefore defend open standards, interoperable protocols, and documented formats, while further distributing forges, mirrors, maintainers, and infrastructures. An open Internet will not survive if it depends on three major American companies.
You have often defended the idea that Europe now has the technical skills necessary to offer credible alternatives to major American platforms. Yet, despite the progress made by actors like Scaleway, OVHcloud, or Clever Cloud, major European groups continue to rely massively on AWS, Microsoft Azure, or Google Cloud. In your view, what is the main obstacle today to the emergence of European actors capable of durably rivaling these global giants?
The main problem is no longer technical competence. It is that Europe does not buy European. Companies and administrations talk about sovereignty, then almost automatically choose AWS, Azure, or Google Cloud. They cite risk, functional richness, habit, or ease of purchase. In the short term, this choice may seem rational. On a continental scale, it is a lack of courage and an absurd industrial policy. We cannot ask Scaleway, OVHcloud, or Clever Cloud to rival groups supported for twenty years by their domestic market if major European clients refuse to entrust them with significant workloads. European actors will progress with demanding clients, volumes, capital, and time. Not with speeches. To make European giants emerge, Europe must begin by accepting to be their first major market.
