Google Cloud's Confidential G4 Instances with NVIDIA Blackwell GPUs are Now Generally Available

Confidential G4 General Availability

Google Cloud has announced the general availability of its Confidential G4 instances. These instances integrate NVIDIA Confidential Computing directly into Google Cloud's AI Hypercomputer, providing enhanced security for processing sensitive data and intellectual property in real-time AI workloads.

Hardware and Availability

The Confidential G4 instances are enabled by NVIDIA RTX PRO 6000 Blackwell Server edition GPUs. They are now available across 36 zones globally, supporting multiple consumption models on both Compute Engine Virtual Machines and Google Kubernetes Engine (GKE) nodes.

Additionally, support for Confidential Space, which facilitates secure multi-party data sharing, is expected to reach general availability in the coming weeks.

Security for AI Workloads

Confidential G4 instances aim to protect data while it is actively being computed without sacrificing GPU throughput or operational agility. This is achieved through a combination of non-exportable hardware keys, verifiable remote attestation, and transparent memory encryption. This security framework helps organizations meet regulatory and compliance requirements for agentic and inference workloads.

Industry Perspective

Taher Behbehani, Founder and CEO of Confidential Core AI, noted that these VMs provide a foundation for privacy-preserving AI and AI operations. He stated that Confidential Core AI builds on Google Cloud to translate enterprise policies into real-time safeguards for AI agents, models, data, and tools, enabling scalable and trusted AI on Google Cloud.