Gradium raises $100 million to expand voice AI in the US

Gradium said it has reopened its seed round to new investors, including Nvidia, lifting the Paris startup’s total seed funding to $100 million and paving the way for a Bay Area office. The move puts fresh capital behind a company betting that the next major AI interface shift will be voice, not just text or images.

The company made the announcement on July 9, 2026, and framed the new money as fuel for hiring and expansion in the United States, where the deepest pools of AI talent, compute and investor attention remain concentrated in the San Francisco Bay Area. For Nvidia, the investment signals confidence that low-latency speech systems will become a core layer of the AI stack, not a niche feature.

Gradium was founded in September 2025 by Neil Zeghidour, Olivier Teboul, Laurent Mazaré and Alexandre Défossez, and it came out of stealth on December 2, 2025 with a $70 million seed round led by FirstMark Capital and Eurazeo. That first financing also drew DST Global Partners, Eric Schmidt, Xavier Niel, Rodolphe Saadé, Korelya Capital and Amplify Partners, putting some of France’s best-known tech backers alongside U.S. venture firms from the start.

The startup says it was spun out of Kyutai, the Paris-based non-profit AI research lab launched in November 2023 by Iliad, CMA CGM and Schmidt Futures. Gradium says the company generated its first revenue within weeks of founding and already has early adopters in gaming, AI agents, customer care, language learning and healthcare, a spread that suggests investors see voice as a horizontal product layer rather than a single-use tool.

Gradium’s platform bundles text-to-speech, speech-to-text and voice cloning through one API, with support for English, French, German, Spanish and Portuguese and more languages in development. Neil Zeghidour has argued that voice AI is still earlier in its evolution than chatbots were before large language models changed the market, and that Gradium’s goal is to remove the trade-offs among quality, scale and latency so voice can become the primary interface between people and machines.

That pitch lands in a crowded field. Gradium is up against ElevenLabs, OpenAI, Anthropic, Meta, Mistral and a large open-source ecosystem, all chasing the same shift toward conversational, real-time software. The commercial opportunity is large, but so are the risks: voice cloning can lower the cost of fraud, and a race to own the interface could concentrate more control in a handful of platform companies.