Nvidia Backs Paris-Based Voice AI Startup Gradium in $100M Seed Round

Paris-based Gradium has successfully closed a massive $100 million seed funding round, bolstered by a strategic investment from Nvidia. The startup is positioning itself to redefine the frontier of human-computer interaction through ultra-low latency audio models.

Solving the Latency Gap in Conversational AI

The primary technical challenge in the current generative AI landscape is the "awkward pause"—the delay between a user's prompt and an AI's vocal response. Gradium is specifically engineering audio models designed to deliver voice at scale with ultra-low latency. By minimizing this response lag, Gradium aims to facilitate natural, fluid, and near-instantaneous conversations, a critical requirement for the next generation of autonomous AI agents.

This technical focus addresses a major pain point for developers building voice-first applications, where even a few hundred milliseconds of delay can break the illusion of human-like interaction and disrupt user experience.

A Pedigree of Deep Learning Expertise

Gradium’s rapid ascent is fueled by a high-caliber founding team with deep roots in the research community. The company was spun out of the prominent French AI lab Kyutai and co-founded by Neil Zeghidour. Zeghidour brings an elite technical pedigree to the venture, having previously contributed to groundbreaking work at Google Brain, DeepMind, and Facebook.

The company's initial $70 million launch in December has now expanded to $100 million following the reopening of the seed round. This capital influx includes heavyweights such as FirstMark Capital, Eurazeo, DST Global Partners, Eric Schmidt, and French telecom billionaire Xavier Niel. The addition of Nvidia as a backer signals significant industry confidence in Gradium's underlying architecture.

Expansion into the Bay Area and Market Competition

While Paris remains a major European hub for AI development, Gradium is making a decisive move to compete on the global stage. The company has announced plans to open an office in the San Francisco Bay Area to tap into the world’s densest concentration of AI talent. This expansion places Gradium in direct proximity to industry titans like OpenAI, Anthropic, Google, and Meta.

The competitive landscape for voice AI is fierce. Gradium faces established leaders like ElevenLabs, which reached an $11 billion valuation earlier this year, as well as integrated voice capabilities from Google’s Gemini. However, Gradium has already demonstrated commercial viability, securing high-profile customers such as the French automotive manufacturer Renault.

Why This Matters for the AI Ecosystem

Gradium’s massive seed round underscores a shift in AI investment: the move from general-purpose LLMs toward highly specialized, modality-specific models. As the industry moves toward "Agentic AI"—systems that can act and communicate independently—the ability to process and generate audio with sub-second latency becomes a foundational necessity rather than a luxury.

Key Takeaways

  • Massive Capitalization: Gradium secured $100 million in seed funding, with participation from Nvidia and high-profile investors like Eric Schmidt and Xavier Niel.
  • Technical Specialization: The startup is focusing on ultra-low latency audio models to eliminate response delays in conversational AI agents.
  • Strategic Global Expansion: To compete for top-tier talent, Gradium is expanding from its Paris roots into the San Francisco Bay Area ecosystem.