NVIDIA and Google Cloud Push Boundaries of AI with Rubin GPUs

Caroline Bishop Apr 22, 2026 12:58

NVIDIA's Vera Rubin GPUs and Google Cloud's AI infrastructure pave the way for agentic and physical AI at unprecedented scales.

NVIDIA and Google Cloud Push Boundaries of AI with Rubin GPUs

NVIDIA (NASDAQ: NVDA) and Google Cloud have unveiled a groundbreaking expansion of their decade-long partnership, targeting the next generation of AI applications. Announced at Google Cloud Next 2026 in Las Vegas, the collaboration revolves around NVIDIA’s Vera Rubin-powered A5X instances and Google Cloud’s AI infrastructure, enabling the creation of "AI factories" capable of scaling to nearly 1 million Rubin GPUs.

These AI factories aim to advance agentic and physical AI—systems designed to reason, plan, and act autonomously—while cutting costs and energy use. NVIDIA’s Rubin GPU platform, part of the cutting-edge Blackwell architecture, offers up to four times the inference throughput of its predecessor, the H100, and consumes 25 times less energy for trillion-parameter models. This makes it a crucial leap forward in reducing the cost and environmental impact of AI workloads.

Rubin GPUs at Scale

The A5X instances, powered by NVIDIA Vera Rubin NVL72 rack-scale systems, promise a 10x reduction in inference cost per token and a 10x increase in token throughput per megawatt compared to earlier models. These instances can scale up to 80,000 Rubin GPUs in a single-site cluster and 960,000 GPUs across multisite clusters, offering unparalleled performance for enterprise-scale AI tasks such as training large language models and running real-time inference workloads.

"The next decade of AI will be shaped by the ability to run demanding workloads on integrated, AI-optimized infrastructure," said Mark Lohmeyer, VP and GM of AI and computing infrastructure at Google Cloud. By combining Google Cloud's scalable infrastructure with NVIDIA's Blackwell-based systems, the platform gives enterprises a flexible and cost-effective solution for complex AI workflows.

Agentic AI and Physical Applications

Beyond infrastructure, the partnership introduces enhancements to NVIDIA’s Gemini Enterprise Agent Platform. Developers can leverage NVIDIA Nemotron and NeMo frameworks to build agentic AI systems capable of managing workflows, simulating environments, and powering robotics. For example, NVIDIA's open models and APIs are now optimized to run in Google Cloud environments, enabling faster training and deployment of agentic AI.

Applications range from cybersecurity tools by CrowdStrike, which fine-tune models for domain-specific threat detection, to industrial solutions like NVIDIA Omniverse and Isaac Sim, which are used to create digital twins and robotics simulations. These tools streamline the transition from design to real-world deployment, spanning industries such as automotive, aerospace, and heavy machinery.

Competitive Edge in AI Hardware

NVIDIA’s Blackwell GPUs bring significant advancements in memory bandwidth, energy efficiency, and tensor performance. These features are crucial for training enormous models and running real-time AI inference. With cloud rental rates for Blackwell GPUs ranging from $2.65/hr to $14.24/hr and complete DGX B200 systems priced at over $500,000, NVIDIA is solidifying its position as the AI hardware leader.

This dominance is reflected in NVIDIA’s current market capitalization of $4.858 trillion, making it the world’s most valuable company. The Vera Rubin platform, with its combination of Rubin GPUs and Vera CPUs, is expected to generate substantial growth opportunities as cloud providers like AWS, Google Cloud, and Microsoft deploy Rubin-based instances throughout 2026.

Path Forward

This latest collaboration underscores NVIDIA's and Google Cloud's commitment to accelerating the deployment of AI at massive scales. With agentic and physical AI poised to transform industries, enterprises now have access to the tools and infrastructure needed to move from prototyping to production seamlessly. As more developers join the ecosystem—over 90,000 in the past year alone—the potential for innovation is enormous.

For traders and investors, the AI hardware race is a key area to watch. NVIDIA’s ability to maintain its technological lead with products like the Rubin and Blackwell platforms strengthens its growth outlook, while its partnership with Google Cloud adds another layer of defensibility against competitors.

The companies' joint efforts will be showcased in upcoming workshops and demos at Google Cloud Next. With enterprise adoption accelerating, NVIDIA continues to set the standard for the AI era.

Image source: Shutterstock