Paris-based voice-AI startup Gradium has reopened its seed round to bring in Nvidia and other backers, pushing the total raised to $100M, the company said Thursday. The fresh capital sits on top of the $70M it disclosed in December when it emerged from stealth, and it will fund a new office in the Bay Area aimed at competing for AI talent alongside Anthropic, Google, Meta, and OpenAI.
The decision to cross the Atlantic is notable given Paris's standing as Europe's most active AI hub. Gradium framed the move as a way to sit closer to the customer base and the engineering pool that the largest US labs draw from. It is a rare acknowledgment from a well-funded European startup that proximity to Silicon Valley still matters, even with Mistral, Hugging Face, and Kyutai anchoring the French scene.
Gradium builds audio models designed to deliver voice at scale with ultra-low latency — the kind of near-instant response that removes the awkward pause common in AI agent conversations. That latency gap is where voice startups are competing hardest right now, because it determines whether an AI voice agent feels usable in a call center, a car, or a consumer app.
Key facts
- 01Gradium reopened its seed round to bring in Nvidia and other investors, taking the total to $100M from an initial $70M raised in December.
- 02The startup is opening a Bay Area office to recruit talent closer to Anthropic, Google, Meta, and OpenAI.
- 03Gradium was spun out of French AI lab Kyutai and co-founded by Neil Zeghidour, formerly of Google Brain, DeepMind, and Facebook.
- 04French automaker Renault is among the customers Gradium has landed since launching out of stealth in December.
- 05The company competes with ElevenLabs, valued at $11B in February, and voice features from Google's Gemini.
The company was spun out of Kyutai, the French AI lab backed by telecom billionaire Xavier Niel, and co-founded by Neil Zeghidour, a researcher who previously worked at Google Brain, DeepMind, and Facebook. The original December round brought in FirstMark Capital, Eurazeo, DST Global Partners, and Eric Schmidt alongside Niel. Nvidia's addition slots the startup into the chipmaker's expanding roster of strategic bets on model builders.
Since its December debut, Gradium has landed French auto manufacturer Renault as a customer — a marquee reference in a market where enterprise voice deployments are still in early innings. Automotive is a natural fit for low-latency voice, given the in-cabin assistant use case and the tolerance for custom integration work.
The competitive field is crowded. ElevenLabs was valued at $11B in February and remains the standard-setter for expressive voice cloning and multilingual output. Google's Gemini ships native voice capabilities, and OpenAI's real-time voice API has become a default for developers building conversational agents. Gradium's pitch is that dedicated audio-model architecture beats a general-purpose model bolted onto a text stack, particularly on latency.
The round also lands in the middle of a busy week for AI infrastructure and tools funding. Ollama, the popular open-model runner, raised $65M as its user base grew to nearly 9M, and Mercor is in talks for a $20B valuation. Capital is flowing to the layer between raw models and end-user applications — inference tools, voice, agent orchestration — where product differentiation is still available.
Skeptics will point out that voice is a feature the frontier labs can commoditize with each model release. Google, OpenAI, and Anthropic can ship native voice at zero marginal cost to customers already paying for their APIs, which puts pressure on pure-play voice startups to either win on latency and quality by a clear margin or lock in vertical customers with integration depth. Gradium has not published latency benchmarks against ElevenLabs or Gemini, and the Renault win is a single reference rather than a repeatable enterprise motion.
The Bay Area expansion is the real signal from this round. European AI companies have historically stayed close to home and relied on remote hiring for US customer coverage. Gradium's willingness to plant a flag in San Francisco — with Nvidia's capital and endorsement behind it — suggests that at the voice-AI layer, the geography of talent still trumps the geography of policy and price. Whether that gamble pays off depends on how much of the voice market the frontier labs decide to keep for themselves.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
