NVIDIA and ElevenLabs Expand Collaboration to Bring Voice Agents to Enterprise and Government

NVIDIA and ElevenLabs Expand Collaboration to Bring Voice Agents to Enterprise and Government

NVIDIA Nemotron models are being added to ElevenLabs’ ElevenAgents platform, giving organisations more options for building and deploying voice agents at scale.

Picture a government helpline where a voice agent handles thousands of simultaneous calls, routes queries accurately, and hands off to a human adviser when needed. That’s the kind of deployment NVIDIA and ElevenLabs are now actively targeting together.

The two companies announced an expanded collaboration at the ElevenLabs New York Summit, with NVIDIA’s Nemotron 3.5 Super VL model now available inside ElevenLabs’ ElevenAgents platform. ElevenAgents is built to help organisations design and deploy voice and chat agents — the kind of automated assistants that can carry on a conversation, reason through a problem, use external tools, and complete multi-step tasks without handing off to a human at every turn.

NVIDIA posted on X that the move is about giving customers more options. “With NVIDIA Nemotron models now in ElevenAgents, organizations have more choice in how they build and customize voice agents,” the company wrote. And the ambition here is clearly large-scale. NVIDIA also posted: “We’re expanding our collaboration with @elevenlabs to bring voice agents to enterprises and governments worldwide.”

What Nemotron 3.5 Super VL Actually Does

Nemotron is NVIDIA’s family of open models built for reasoning-heavy, agentic work — the kind of tasks that require a system to think through several steps, call on other tools or data sources, and arrive at a useful answer rather than just pattern-matching a response.

ElevenLabs says Nemotron 3.5 Super VL achieved leading end-to-end task-completion accuracy on its own customer-support benchmark. The company also claims costs around 70% lower than comparable leading open models, though that figure comes from ElevenLabs itself and hasn’t been independently verified. When reasoning is switched off, ElevenLabs reports it as one of the three fastest models on the platform.

Speed and cost matter enormously when you’re running voice agents at national scale. A fraction of a second’s delay in a spoken conversation feels jarring to a caller. And if you’re fielding millions of interactions, the cost per call adds up fast.

Infrastructure Behind the Partnership

The collaboration isn’t just about which model sits inside the platform. ElevenLabs says it’s adding 15 megawatts of NVIDIA compute capacity, using Blackwell Ultra systems and NVIDIA’s Vera Rubin platform. That’s a sizeable infrastructure commitment — Blackwell Ultra is NVIDIA’s current high-performance data centre chip generation, and Vera Rubin is the next-generation platform following it.

So this isn’t a software-only deal. ElevenLabs is betting heavily on NVIDIA hardware to power the deployments it’s promising.

One concrete example the companies have announced: a joint project to build an on-premises voice agent for Poland’s national e-Health platform. That’s a real public-sector deployment, handling health information for citizens, and it gives a sense of the kind of government-scale work both companies are now pursuing.

Why Enterprises and Governments Are Paying Attention

Mati Staniszewski, chief executive officer of ElevenLabs, said: “With NVIDIA Nemotron reasoning inside ElevenAgents and NVIDIA accelerated computing powering large-scale deployments, enterprises and governments can run voice agents securely at national scale — and make talking the fastest way to get things done.”

Staniszewski has also described voice agents as becoming “critical infrastructure” — a framing that reflects how seriously some in the industry now view this technology.

But that framing also raises legitimate questions. If voice agents become infrastructure, then failures, inaccuracies, or security breaches carry real consequences for real people. Privacy is an obvious concern — voice interactions with government services involve sensitive personal data. Accessibility is another: not every resident or service user will find a voice interface easy to use, and a clear route to a human adviser remains essential. Accountability matters too. When an automated system gives someone wrong information about a benefit, a health service, or a legal entitlement, who is responsible?

None of those questions have easy answers, and neither company addressed them directly in the announcement.

What Comes Next

The Poland e-Health project is the most concrete near-term deployment named so far. Beyond that, both companies have signalled that enterprise and government customers across multiple countries are the target market. The involvement of NVIDIA’s Kari Briski alongside ElevenLabs representatives in the summit discussion suggests this is being treated as a strategic partnership rather than a one-off product integration.

The details reported here reflect what was announced at the ElevenLabs New York Summit. The precise date of that announcement has not been confirmed and is not stated here.

What This Means for Kent Residents

No confirmed deployments involving Kent County Council, Medway Council, or NHS Kent and Medway ICB have been identified. However, the broader direction of travel — NVIDIA and ElevenLabs actively courting government customers — means UK public bodies, including those in Kent, may eventually be evaluating this kind of voice-agent technology for services ranging from council enquiry lines to health information platforms. Any such adoption in the public sector would need to meet UK data-protection requirements and ensure residents can still reach a human when they need one.

Source: @nvidia

⚡

NVIDIA and ElevenLabs Expand Collaboration to Bring Voice Agents to Enterprise and Government Quiz

5 questions