Zoom Turns to Nvidia’s Nemotron to Power Its AI Companion and New 49B-Parameter LLM

Zoom has partnered with Nvidia to add the chipmaker’s Nemotron technology to its federated AI architecture. The integration aims to strengthen Zoom’s AI Companion, enabling it to support better those organizations in finance, healthcare, and government—all regulated spaces. In addition, Nemotron is being used to fine-tune a new 49-billion-parameter Large Language Model (LLM).

“With the help of Nvidia Nemotron open technologies, we’re accelerating the development of our enterprise retrieval-augmented generation (RAG) capabilities, allowing AI Companion to work seamlessly with Microsoft 365, Microsoft Teams, Google Workspace, Slack, Salesforce, and ServiceNow,” X.D. Huang, Zoom’s chief technology officer, remarks in a release. “This partnership allows us to deliver powerful, security-focused, and scalable AI experiences to our customers at rapid speed.”

Because Zoom’s platform has a federated architecture, multiple models can be involved. Queries will be routed between not only the video communication firm’s Small Language Models (SLMs) and other LLMs, including now Nvidia’s Nemotron. It’s believed that adopting this approach enables the system to react faster, with SLMs addressing smaller tasks while LLMs, which are likely equipped with complex reasoning, handle more complex issues. In addition, data doesn’t get routed to a central server, which can be a concern for those in regulated industries.

Zoom announced the third-generation version of its AI Companion last month. The company boasted that this platform would eliminate the friction from modern work. To do so, it includes an agentic AI assistant that proactively helps users turn meetings, chats, and tasks into actionable outcomes by providing insights, note-taking, calendar optimization, writing support, real-time translations, and team-focused automation.

By integrating Nvidia’s Nemotron, AI Companion 3.0 will receive a boost in speed, reasoning, and efficiency. This means enterprises can get real-time meeting transcriptions, instant summarizations, and accurate multi-step workflow assistance more efficiently—no delays or high computing costs. This will be important for those in finance, healthcare, government, and other similar fields.

Kari Briski, Nvidia’s vice president of generative AI software, shared that “the integration of Nvidia Nemotron in Zoom AI Companion allows enterprises…to experience a private, powerful, and personalized work environment for enhanced productivity. Our collaboration with Zoom to bring smart and efficient AI reasoning to customers is just beginning.”

But that’s not all.

Zoom reveals it has a new, unnamed LLM based on Nvidia’s Nemotron. Built using Nvidia’s NeMo tools, the model contains 49 billion parameters—placing it in the upper midrange of today’s LLM. It’s designed to maximize performance for enterprise use cases, balancing speed, cost, and accuracy. Sadly, more specific details about the Zoom LLM, such as benchmarks or whether this is an open or closed model, weren’t immediately disclosed. I’ve reached out to Zoom with additional questions and will update if I receive a response.

Featured Image: Credit: Adobe Firefly