Google Cloud Expands Nvidia Partnership With New G4 Virtual Machines

Google Cloud (GOOG, Financials) is deepening its partnership with Nvidia (NVDA, Financials), launching a new line of G4 virtual machines built to power advanced artificial intelligence and industrial applications.

The G4 VMs run on Nvidia's RTX PRO 6000 Blackwell Server Edition GPUs and AMD's EPYC Turin CPUs, combining high-performance graphics and processing for enterprise-level AI work. The company said the new setup delivers up to nine times the throughput of its previous G2 instances.

Google said the upgraded machines will support everything from multimodal AI inference and 3D visualization to robotics simulations and generative design.They're meant to speed up things like generating images from text and fine-tuning huge language models, which are both areas where computational demand is rising quickly.

Nvidia also said that you can now get its Omniverse and Isaac Sim platforms through the Google Cloud Marketplace. Using digital twins and AI-driven settings, these solutions let manufacturers, automakers, and logistics businesses test out how things work in the real world.

This release completes our full stack built on the Blackwell platform, Nvidia said, referencing a lineup that now spans large-scale AI training systems like the GB200 NVL72 and HGX B200, down to G4-based inference and visualization machines.

The launch underscores how Google and Nvidia are positioning themselves to capture demand from businesses scaling AI infrastructure from generative models to industrial automation.