Friday, June 26, 2026
EN·DarkSubscribe
AI Infrastructure · News & Analysis
HomeData CentersReport
Data Centers · Report

Nvidia's Rubin GPU lands in Google's virtual machines, enabling multi-site clusters approaching 1 million GPUs.

Hyperscaler adoption of cutting-edge GPU architecture at record scale validates infrastructure expansion requirements and signals continued capacity growth.
Trade pressSlicast · April 27, 2026 · Global · Source: wccftech.com
importance 82

Google and NVIDIA have announced a partnership providing access to as many as one million NVIDIA GPUs to power the newly launched A5X instances. The A5X instances are Google's latest products designed specifically to run agentic artificial intelligence workloads and are part of Google's AI Hypercomputer portfolio, which also powers the firm's Gemini platform and its consumer and enterprise AI offerings. These instances rely on NVIDIA's network accelerators that enable the development of single and multi-cluster computing infrastructure for AI workloads.

The A5X instances represent Google's first offering designed to work with NVIDIA's latest Vera Rubin AI GPUs. According to the announcement, the A5X will use NVIDIA's ConnectX-9 NICs, which are designed to accelerate AI workloads in cloud infrastructure run on ethernet. These new capabilities are specifically designed for agentic AI workloads, which rely on a group of AI agents to take a piece-wise approach to solving problems or tasks.

Google's Virgo platform plays a central role in enabling this infrastructure by connecting multiple AI chips within a single data center and across multiple sites. The ConnectX-9 NICs, along with Google's Virgo platform, will allow users to access as many as 80,000 Rubin GPUs in a single cluster and 960,000 GPUs in a multisite cluster. Virgo also supports Google's tensor processing units (TPUs), and can connect as many as 134,000 TPUs in a single data center and more than a million chips across multiple sites.

According to NVIDIA, the A5X instance is capable of delivering 10x lower inference costs per token and 10x higher throughput per megawatt compared to the previous generation. NVIDIA also highlighted physical and industrial AI applications, noting that products from firms such as Cadence and Siemens are powered through its infrastructure and available on Google Cloud. The announcement includes that Google's Gemini platform can deploy agentic models and workflows across industries such as cybersecurity.

Read the original
Nvidia's Rubin GPU lands in Google's virtual… · Slicast