Vultr deploys NVIDIA Nemotron 3 Nano Omni in cloud

Vultr has deployed NVIDIA Nemotron 3 Nano Omni on its cloud platform, expanding the companies' existing collaboration in artificial intelligence infrastructure.

The model is designed for multimodal enterprise agent systems, allowing software agents to work across audio, video, images, documents and text. Customers can deploy it on dedicated NVIDIA GPU clusters or access it through Vultr's serverless inference service, which uses NVIDIA Dynamo 1.0.

Nemotron 3 Nano Omni is part of NVIDIA's Nemotron 3 family of open models. According to Vultr, the open model structure gives developers more control over how they configure and run systems across different environments while reducing dependence on a single vendor setup.

The deployment adds to Vultr's push to attract developers and companies building so-called agentic AI systems, in which software agents handle tasks with limited human intervention. Here, the focus is on sub-agents that can interpret several types of data rather than only text.

Vultr has been building its position in the cloud-based AI infrastructure market by offering access to graphics processing units for model deployment and inference. The latest NVIDIA model is available through both dedicated clusters and a serverless option, giving customers two ways to run workloads depending on their requirements.

J.J. Kardwell, chief executive officer of Vultr, said the rollout is intended to support customers building AI agent projects.

"We have embraced NVIDIA Nemotron to reinvent enterprise AI inference for our customers, and we're thrilled to deliver the highly efficient and accurate Nemotron 3 Nano Omni model to improve the performance of their agentic projects," Kardwell said.

He added: "We're committed to providing high-performance infrastructure for agentic AI, and Nemotron 3 Nano Omni offers a powerful boost to these capabilities."

Model access

The launch reflects a broader industry shift toward multimodal AI models that can process different forms of input within the same system. For businesses, that can include workflows in which an agent reads documents, interprets images, analyses spoken content and responds to text prompts as part of a single task chain.

For cloud providers, support for these models is becoming a competitive differentiator as enterprise customers weigh where to host and manage AI applications. Companies are also under pressure to limit operational complexity and avoid becoming too tied to a single software stack or infrastructure supplier.

Vultr describes itself as an NVIDIA Preferred Partner and Exemplar Cloud. Its infrastructure is used by customers in 33 cloud data centre regions across six continents, giving it a global footprint for AI and cloud computing workloads.

The company is privately held and has sought to differentiate itself from larger public cloud rivals by focusing on simpler cloud services and broad geographic coverage. In corporate background material, it says it serves hundreds of thousands of active customers across 185 countries.

Vultr has also drawn attention for its valuation. The company said it announced an equity financing in December 2024 that valued the business at $3.5 billion, underscoring investor interest in infrastructure groups linked to AI demand.

NVIDIA tie-up

The deployment also deepens Vultr's commercial relationship with NVIDIA, whose chips and software have become central to the current AI build-out. Vultr plans to expand its NVIDIA Dynamo deployment into systems based on the next-generation NVIDIA Vera Rubin platform later this year.

Amanda Saunders, director of generative AI software at NVIDIA, said the company is seeing demand rise as enterprises assign more complex work to AI agents.

"As enterprise agents start taking on more complex, multimodal tasks, teams need models and infrastructure that can scale seamlessly as people realize how much productivity they gain with AI agents," Saunders said.

She added: "With Nemotron 3 Omni Nano now available on Vultr, developers can build and launch agent systems that keep up as enterprises build agents to power the future of work."

The announcement highlights how cloud infrastructure providers are trying to combine access to sought-after AI models with the computing resources needed to run them. In the current market, providers that can offer both model access and GPU supply are aiming to secure a larger share of enterprise AI spending.

Vultr said Nemotron 3 Nano Omni is intended to help enterprise agents complete tasks across multiple data types while giving developers direct access through its existing cloud services.