France is currently locked in a high-stakes race to maintain technological sovereignty in an era dominated by Silicon Valley and Beijing. While the global conversation often centers on the behemoths of San Francisco, a new vanguard of European talent is attempting to rewrite the rules of artificial intelligence from Paris. Among them is Neil Zeghidour, a 35-year-old AI researcher whose recent inclusion in Le Nouvel Obs’ prestigious list of the “50 who will make tomorrow” signals a shift in how France views its digital future.
Zeghidour is not merely another engineer in the AI gold rush; he represents a specific hybrid of academic rigor and industrial scale. As a founding member of the Kyutai research lab and the director of the startup Gradium, he is targeting the most human element of computing: the voice. By moving beyond the clunky “text-to-speech” pipelines that define current virtual assistants, Zeghidour is building systems that can hear, feel and respond in real-time, mirroring the fluid cadence of human conversation.
His recognition by Le Nouvel Obs places him within a curated cohort of under-40s tasked with solving the planet’s most pressing crises—from ecology to societal equity. For Zeghidour, the mission is a balance of open-source discovery and commercial viability, aiming to ensure that the next generation of AI is not just powerful, but responsible and accessible.
The Moshi Breakthrough: Beyond the Text Box
At the heart of Zeghidour’s current impact is “Moshi,” a conversational AI developed at the Kyutai lab. Unlike traditional AI voice modes—which typically transcribe speech to text, process the text, and then convert a response back to audio—Moshi is designed for native speech-to-speech interaction. This architectural shift drastically reduces latency, allowing for a “real-time” experience that can handle interruptions and emotional nuances.
Zeghidour’s ambition is to create an interface that transcends the written word entirely. He envisions an AI capable of reproducing specific historical or public voices—such as those of Donald Trump or General Charles de Gaulle—while seamlessly switching between languages or transcribing oral data on the fly. According to reportage by Le Nouvel Obs, Moshi has already seen significant traction, recording over one million downloads per month and becoming a foundational tool for other AI startups.
This capability directly challenges the dominance of OpenAI’s GPT-4o and Google’s Gemini, suggesting that a lean, European research model can compete with the massive compute clusters of the American giants.
From the ENS to the Meta AI Lab
Zeghidour’s trajectory is a testament to the French “Grandes Écoles” system, tempered by a global perspective. Raised in Paris in a household where academic achievement was a core value—his mother a microbiologist and his father, Slimane Zeghidour, a seasoned journalist for TV5 Monde—Neil initially aspired to follow his father into journalism.
However, a pivot toward mathematics led him to Université Dauphine and the École normale supérieure (ENS), the pinnacle of French academic prestige. His career then moved through the financial sector at Société Générale before he joined the AI laboratory at Meta (formerly Facebook). It was here that he honed the skills necessary to scale complex models, bridging the gap between theoretical mathematics and consumer-facing software.
This journey from the halls of the ENS to the corridors of Big Tech provided him with a unique vantage point: he understands the raw power of American corporate AI, but remains committed to the European ideal of research for the public quality.
The Dual Path: Kyutai and Gradium
One of the most distinct aspects of Zeghidour’s current work is the strategic separation between his research and his commercial interests. He operates across two distinct entities with different mandates:
- Kyutai: A non-profit research laboratory dedicated to the open exploration of AI. Zeghidour maintains that Kyutai’s mission is to remain a sanctuary for research, focusing on discovery rather than the commercialization of its findings.
- Gradium: A commercial startup where Zeghidour applies AI breakthroughs to scalable products. This venture has already attracted significant capital, reportedly raising €60 million in 2025 from a consortium of prestigious French and American investors.
This “dual-track” approach allows him to contribute to the global scientific community while simultaneously building a sustainable business capable of surviving in a market dominated by trillion-dollar companies.
| Feature | Traditional AI Voice (Pipeline) | Moshi (Native Speech-to-Speech) |
|---|---|---|
| Process | Speech $rightarrow$ Text $rightarrow$ LLM $rightarrow$ Text $rightarrow$ Speech | Direct Audio $rightarrow$ Audio |
| Latency | Higher (noticeable lag) | Ultra-low (near-instant) |
| Nuance | Often robotic or monotone | Capable of emotional inflection |
| Interaction | Turn-based (Wait for response) | Fluid (Allows interruptions) |
The Broader Stakes for European AI
Zeghidour’s success is more than a personal achievement; it is a case study in France’s broader strategy to avoid “digital colonization.” By fostering talent that can navigate both the academic and venture capital worlds, France is attempting to build a sovereign AI ecosystem that prioritizes transparency and ethical deployment over pure profit.
The challenge remains the “compute gap”—the sheer amount of hardware required to train these models. However, the €60 million injection into Gradium suggests that investors believe the European approach—focused on efficiency and specialized architecture rather than brute-force scaling—may be the winning strategy.
Disclaimer: This article mentions venture capital funding and startup investments. This information is for editorial purposes and does not constitute financial advice.
As Kyutai continues to release its research and Gradium scales its operations, the next critical milestone will be the integration of Moshi’s architecture into wider consumer applications. The industry is now watching to see if these French-led innovations can move from the “list of those who will make tomorrow” to the standard tools of today.
What do you think about the rise of European AI? Share your thoughts in the comments below or share this story on social media.
