NVIDIA and Google Cloud Partner to Bring Agentic AI On-Premises for Enterprises

NVIDIA announced a collaboration with Google Cloud to make Google’s Gemini AI models available on premises through Google Distributed Cloud, powered by NVIDIA’s Blackwell HGX and DGX platforms with Confidential Computing. This move brings secure, high-performance agentic AI reasoning to enterprises that need to comply with strict data privacy and sovereignty requirements.

With NVIDIA Confidential Computing, enterprises can protect both sensitive user prompts and fine-tuning data, ensuring they are not accessible or alterable. At the same time, model providers can prevent unauthorized tampering, creating a dual-layer security solution.

Agentic AI, which enables systems to autonomously reason and adapt in dynamic environments, is reshaping industries. Examples include IT support systems that not only identify issues but also resolve them, and financial systems that autonomously respond to fraud risks in real time.

Until now, strict security constraints kept many organizations from using such powerful models on premises. The integration of Gemini models with the NVIDIA Blackwell platform addresses that gap, offering a breakthrough in both performance and compliance.

NVIDIA and Google Cloud are also enhancing deployment observability and scalability through tools like the new GKE Inference Gateway and NVIDIA NeMo Guardrails. Further, NVIDIA Dynamo will soon offer better oversight for agentic AI workloads across distributed environments.

Learn more by exploring sessions and demos from NVIDIA at Google Cloud Next 2025.