Nvidia Pushes Open AI Models for 2,000-Tool Enterprises and Sovereign Users
Updated
Updated · Computerworld · Jul 29
Nvidia Pushes Open AI Models for 2,000-Tool Enterprises and Sovereign Users
1 articles · Updated · Computerworld · Jul 29
Summary
Kari Briski said Nvidia’s open-weight Nemotron models let enterprises and countries inspect, adapt and control AI with their own data, framing openness as key to security, innovation and sovereignty.
2,000 tools in a single enterprise—or 2,000 regional tools in a local market—require locally post-trained models, she said, as AI shifts from question-answering to agentic systems that call tools to complete tasks.
Nvidia is offering Nemotron in Nano, Super and Ultra sizes and through cloud model-as-a-service partners, aiming to let developers start on smaller GPUs and scale up without rebuilding workflows.
In the last 90 days, workloads shifted sharply toward agentic use cases, Briski said, pushing Nvidia to emphasize token efficiency, tracing and open collaboration through the Nemotron Coalition and Open Secure AI Alliance.
For regions with limited compute, Briski cast open models and open data as a bootstrap for sovereign AI, arguing they reduce the need to build a base model from scratch and widen access beyond grant-based research clusters.