NVIDIA (NASDAQ: NVDA) and Google Cloud (NASDAQ: GOOGL) announced expanded collaboration at Google Cloud Next in Las Vegas, introducing new AI infrastructure capabilities including NVIDIA Vera Rubin-powered A5X instances that can scale to nearly 1 million Rubin GPUs.

The partnership introduces A5X bare-metal instances powered by NVIDIA Vera Rubin NVL72 rack-scale systems, which the companies said deliver up to 10 times lower inference cost per token and 10 times higher token throughput per megawatt compared to the previous generation. The A5X instances can scale to 80,000 NVIDIA Rubin GPUs within a single site cluster and up to 960,000 GPUs across multiple sites.

Google Cloud will offer a preview of Google Gemini models running on NVIDIA Blackwell and Blackwell Ultra GPUs through Google Distributed Cloud. The collaboration also includes confidential virtual machines with NVIDIA Blackwell GPUs, marking the first confidential computing offering with Blackwell GPUs in the cloud.

The companies announced that NVIDIA Nemotron 3 Super models are available on Gemini Enterprise Agent Platform. Google Cloud introduced a managed reinforcement learning API built with NVIDIA NeMo RL for training at scale.

"By combining Google Cloud's scalable infrastructure and managed AI services with NVIDIA's industry-leading platforms, systems and software, we're giving customers flexibility to train, tune and serve everything from frontier and open models to agentic and physical AI workloads," said Mark Lohmeyer, vice president and general manager of AI and computing infrastructure at Google Cloud.

OpenAI is using NVIDIA GB300 and GB200 NVL72 systems on Google Cloud for inference workloads including ChatGPT, according to the announcement. CrowdStrike uses NVIDIA NeMo libraries running on the platform for cybersecurity applications.

NVIDIA received Google Cloud Partner of the Year recognition in AI Global Technology Partner and Infrastructure Modernization Compute categories. The companies reported more than 90,000 developers have joined their joint developer community.