At VivaTech, the question seemed simple: What is the best strategy for investing in artificial intelligence? The answer, however, was less straightforward. Around the table, three perspectives converged: that of British sovereign wealth funds, that of the French startup ecosystem, and that of European growth equity. Three ways of looking at the same trend, with one thing in common: AI is no longer just a sector. It’s infrastructure.
Hala Fadel, of Eurazeo, drew a useful line in the sand right from the start. Anything marketed as “analytics with AI” already seems shaky. It’s too easy to replicate, too dependent on existing models, and too close to mere software packaging. The value lies elsewhere: in infrastructure, in the tools that organize data, in the deeper layers of the cloud, but also in what we might call “physical AI”—robotics, sensors, cameras, industrial systems—where data becomes proprietary because it arises from concrete use.
This distinction matters. It separates a passing fad from a barrier to entry. An AI agent connected to an interface can quickly win people over. A system integrated into a production line, an autonomous fleet, or a medical protocol is not so easily replaced. Investors are less interested in the promise of such systems than in their inertia: the inertia of workflows, accumulated data, and exit costs.
The European argument now hinges on this question: Where can the continent build global companies, rather than merely regional alternatives? In April 2026, the United Kingdom launched a 500 million-pound Sovereign AI program to support AI companies with strong British roots, with the explicit goal of making the country a producer of AI, not just a consumer. (GOV.UK)
In Paris, the debate is broader. European technological sovereignty has emerged as one of the central themes of VivaTech 2026, against a backdrop of persistent dependence on U.S. infrastructure: cloud services, chips, and foundational models. Reuters reported on June 17, 2026, that several European leaders viewed the recent restrictions on certain U.S. models as a warning sign regarding the continent’s vulnerability. (Reuters)
Capital, in this context, takes on an almost diplomatic role. It no longer merely finances companies; it buys time, autonomy, and a seat at the table. Susanna, from the British sovereign wealth fund, put it simply: in certain sectors, owning domestic market leaders provides leverage. Without them, Europe remains a customer.
But the continent is moving forward with its usual paradox: it wants to create giants while regulating early on. The AI Act, which took effect in 2024, remains a strong political milestone. It establishes the concept of trustworthy AI, based on risk, transparency, and accountability. But in the room, the disagreement was clear. For Hala Fadel, regulating now amounts to putting the brakes on a technology whose economic contours have not yet stabilized. For Maya, from France Digitale, regulation can also become a positioning tool: positioning Europe as a hub for responsible AI—provided that this narrative does not take precedence over capital, public procurement, and speed of execution.
That may be where the crux of the matter lies. Europe doesn’t just need better business models. It needs customers. It needs public procurement systems capable of buying European products. It needs funds capable of supporting late-stage funding rounds. A continental legal vehicle, such as “EU Inc.,” that would spare founders from having to go through Delaware to secure funding, compensate their talent, and grow without administrative hurdles.
The parallel with the United States is illuminating. SpaceX, often cited as an example of entrepreneurial boldness, also built itself with massive support from U.S. government contracts. Europe sometimes views this model with suspicion, as if helping its leading companies were tantamount to keeping them artificially alive. This is a misinterpretation. In AI, the government does not replace the market; it can help kick-start it.
That leaves the question of valuations. They’re high—sometimes seemingly absurd. But in the seed stage, price isn’t always the best indicator. The best founders raise a lot of money because they need to pay for computing power, attract top talent, and hold their own against competitors backed by U.S.-scale funding. In AI, capital itself becomes a form of moat.
Caution is still warranted, however. A viable company isn’t defined by the use of the word “AI” in its pitch deck. It’s recognized by the quality of its technical team, its relevance to real-world use cases, its customer retention, the depth of its data, and its ability to improve over time. The rest is often just window dressing.
The sectors mentioned already form a hierarchy: healthcare, biotechnology, defense, manufacturing, robotics, cybersecurity, and software productivity. Not because they are more appealing, but because they combine sensitive data, operational complexity, public needs, and measurable productivity gains. In these sectors, AI acts less as an interface and more as a force for transformation.
By 2030, the European AI landscape will be judged less by the number of startups created than by the quality of its exits, its IPOs, and its industry leaders capable of remaining European without staying small. That is the crux of the matter: not to turn AI into yet another polite dependency, but into an industry.
The wave is here. The question is whether Europe wants to ride it, or just regulate the foam.
Cette publication est également disponible en :
