Gradium just emerged from stealth with one of the largest seed rounds in AI voice history. The Paris-based startup, spun out of French AI lab Kyutai, secured $70 million from FirstMark Capital and Eurazeo to build ultra-low latency voice models that respond almost instantly. With backing from telecom billionaire Xavier Niel and former Google CEO Eric Schmidt, Gradium is positioning itself as Europe's answer to OpenAI's voice dominance.
Gradium just pulled off what might be the most impressive stealth exit in AI voice tech this year. The Paris-based startup emerged Tuesday with a staggering $70 million seed round that puts it in direct competition with the voice AI giants before most people even knew it existed.
The round was led by FirstMark Capital and Eurazeo, with a investor lineup that reads like a who's who of tech power players. French telecom billionaire Xavier Niel, who backed the original Kyutai AI lab that spawned Gradium, doubled down on his bet. DST Global Partners and former Google CEO Eric Schmidt also joined the round, signaling serious confidence in founder Neil Zeghidour's vision.
Zeghidour knows voice AI intimately. The Google DeepMind veteran spent years working on voice models before spinning out of Kyutai just three months ago in September. According to TechCrunch's original report, his team has developed audio language models designed to deliver voice at scale with what they call "ultra-low latency" - essentially AI voices that respond almost instantly.
That speed promise comes at a crucial time. While OpenAI dominated headlines with ChatGPT's voice mode and Anthropic launched Claude's voice features, both still struggle with the lag that breaks conversational flow. Users often wait several seconds for responses, creating an awkward pause that reminds you you're talking to a machine.
Gradium's multilingual approach also sets it apart from the Silicon Valley competition. The startup launched with native support for English, French, German, Spanish, and Portuguese, with additional languages coming. That's a direct challenge to the English-first approach most American AI companies have taken, potentially opening up massive European and Latin American markets that have felt underserved.
But Gradium faces serious competition. OpenAI continues refining its voice models, Anthropic just launched voice mode for Claude, and Llama now includes multimodal capabilities. Well-funded startups like have already captured significant developer mindshare, while hundreds of voice models populate Hugging Face.
