Google Cloud has significantly expanded its accelerated computing offerings, making G4 virtual machines generally available. These new VMs are powered by NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, marking a critical step in transforming enterprise workloads from visual computing to advanced agentic and physical AI. This move positions Google Cloud to address the most demanding challenges across various industries.
At the core of the G4 VM lies the NVIDIA RTX PRO 6000 Blackwell Server Edition GPU, engineered as a universal platform for both AI and visual computing. Its design integrates Fifth-Generation Tensor Cores, which deliver a substantial leap in AI performance by supporting new data formats like FP4 for faster processing and reduced memory footprint. Simultaneously, Fourth-Generation RT Cores provide over double the real-time ray-tracing performance compared to the previous generation, enabling cinematic-quality graphics and highly photorealistic simulations. This dual capability makes the Blackwell GPU exceptionally versatile for a wide array of high-performance tasks.
The G4 VMs on Google Cloud are built for massive scale, configurable with up to eight RTX PRO 6000 GPUs, offering a formidable 768 GB of GDDR7 memory. These instances are deeply integrated into Google Cloud’s AI Hypercomputer architecture, ensuring native compatibility with services like Google Kubernetes Engine and Vertex AI. This integration simplifies containerized deployments and streamlines machine learning operations for complex AI workloads. Furthermore, the platform accelerates large-scale data analytics on Apache Spark and Hadoop with Dataproc, while also supporting a broad ecosystem of popular third-party engineering and graphics applications.
Blackwell's Impact on Industrial Digitalization
Beyond raw compute, Google Cloud customers can now leverage NVIDIA Omniverse, a robust collection of libraries and frameworks for developing industrial digitalization applications based on Universal Scene Description (OpenUSD). The availability of Omniverse and NVIDIA Isaac Sim as virtual machine images (VMIs) on the Google Cloud Marketplace is a game-changer for key industries. Enterprises can now build and operate highly accurate digital twins of factories and products, simulating and optimizing operations in real time. This capability, powered by NVIDIA Cosmos and Omniverse Blueprints, extends to accelerating robotics development, allowing AI-driven robots to be trained and validated in physics-based virtual environments before physical deployment.
The expanded NVIDIA Blackwell Google Cloud partnership also brings the full NVIDIA software stack to bear on a range of high-demand workloads. Developers can utilize the NVIDIA Nemotron family of open reasoning models and NVIDIA Blueprints for building sophisticated AI agents. For optimized, high-performance inference, NVIDIA NIM microservices provide enterprise-grade security and support. Scientific and High-Performance Computing benefits from NVIDIA CUDA-X libraries, with core genomics algorithms seeing up to 6.8x faster throughput on the RTX PRO 6000 Blackwell GPU. Even remote creative and design pipelines are enhanced through NVIDIA RTX Virtual Workstation software, delivering high-performance virtual workstation instances from G4 VMs to any device.
This comprehensive announcement establishes an end-to-end platform built on the NVIDIA Blackwell architecture, spanning from the GB200 NVL72 and HGX B200 for massive-scale AI training to the RTX PRO 6000 Blackwell for inference and visual computing. According to the announcement, this unified architecture provides a seamless experience for accelerating every workload, empowering enterprises to tackle complex, multi-stage pipelines from data analytics to physical AI within a single, consistent cloud ecosystem. This strategic alignment solidifies Google Cloud's position as a premier destination for advanced AI, simulation, and industrial digitalization initiatives.



