Gradium raises $100M seed, backed by Nvidia, for ultra-low-latency voice AI models
The AMW Read
Massive seed round for voice AI with Nvidia backing is a significant signal in multimodal/generative media, but the segment already has well-funded incumbents like ElevenLabs.
Gradium raises $100M seed, backed by Nvidia, for ultra-low-latency voice AI models
Paris-based Gradium, a startup spun out of French AI lab Kyutai, has raised a total of $100M in a seed round now backed by Nvidia. Gradium originally launched in December with $70M from investors including FirstMark Capital, Eurazeo, DST Global, Eric Schmidt, and Xavier Niel. The company builds voice AI models that deliver near-instantaneous response, targeting the awkward pause in AI agent conversations. Gradium will use the capital to open a Bay Area office and compete for talent near the US AI hub. Customers include French auto manufacturer Renault.
This massive seed round — nearly unheard-of at that stage — signals a capital-compression pattern where frontier voice AI startups attract hyperscaler investment before proving product-market fit at scale. Nvidia's involvement as a strategic backer reinforces the compute-intensity of real-time voice inference and positions Gradium to leverage Nvidia's distribution ecosystem. The company's origin inside Kyutai, a research lab, mirrors the pattern of talent spinouts carrying foundational research into commercial products.
Gradium now competes directly with ElevenLabs, valued at $11B in February, and Google's Gemini. The Paris-to-Bay Area expansion also highlights a talent-drain dynamic: even well-funded European AI labs feel pressure to plant a flag in Silicon Valley. Gradium's ability to win Renault as a customer suggests early traction in automotive voice interfaces, but the seed-stage valuation and Nvidia alignment raise the bar for execution.
