Google Cloud Sets November GA for Vera Rubin

Google Cloud will make its bare-metal A5X instances, built on NVIDIA's Vera Rubin NVL72 systems, generally available in November, the company said on October 21 at NVIDIA's GTC Berlin conference. Its Virgo network can link 80,000 Rubin GPUs at one site and up to 960,000 across several sites. Ineffable Intelligence, the London startup founded by AlphaGo creator David Silver, is the first customer, as trade outlet Converge Digest noted on October 25.

Key takeaways

• Google Cloud A5X instances on NVIDIA Vera Rubin NVL72 reach general availability in November 2026, with new EMEA regions planned.
• Virgo is designed for 80,000 Rubin GPUs per site and up to 960,000 across sites — design ceilings, not launch capacity.
• Confidential Computing on G4 GPU VMs is now generally available in 36 zones.

What Google Cloud A5X includes

A5X combines Vera Rubin NVL72 racks with NVIDIA ConnectX-9 SuperNICs, Google's Virgo Network and fourth-generation liquid cooling inside its AI Hypercomputer stack. In a Google Cloud blog post, vice presidents Mark Lohmeyer and Ankur Mehrotra said general availability is "coming next month," along with new local regions for EMEA data residency. Google did not name the regions or give prices.

Google Cloud Managed Lustre will support NVIDIA GPUDirect Storage on A5X, a direct RDMA path into GPU memory that bypasses the CPU. Google also described KV caching that moves context beyond GPU high-bandwidth memory for long agent sessions, while GKE Managed DRANET and Kueue handle RDMA networking and topology-aware scheduling.

Ineffable Intelligence and WPP test A5X

According to an NVIDIA blog post, Ineffable Intelligence is using A5X for large-scale reinforcement learning. "We're building superlearners that continually improve through their own experience," Silver said.

Advertising group WPP built a creative-agent reference architecture on A5X with NVIDIA Nemotron models and NVIDIA OpenShell. NVIDIA said that in initial tests, workloads that usually needed a multi-node cluster ran on a single A5X instance, with a human approving each job. "The true promise of agentic AI isn't just speed, it's freeing up human creativity from the friction of modern production pipelines," said Stephan Pretorius, WPP's chief technology officer.

Confidential GPU computing on G4

Google made Confidential Computing generally available on G4 virtual machines with NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs. The G4 announcement reports less than 3.5% throughput loss versus standard G4 in Google's inference tests. Customer Vertebrae reported a 5.2% rise in end-to-end latency in exchange for hardware-level privacy. Google also introduced Confidential Bare Metal on GB300-based A4X Max instances, without a release date.

Context: Nvidia racks and power limits

Google builds its own TPUs but remains a major buyer of NVIDIA hardware. Vera Rubin NVL72 appeared in MLPerf in September only as a preview system, and hyperscaler orders now run into the millions of GPUs, as Amazon's tripling of its NVIDIA commitment to two million GPUs showed. The companies said they are exploring NVIDIA DSX MaxLPS to raise performance per watt within existing data center power budgets.

What's next

The November launch should reveal how much A5X capacity Google offers, where and at what price. Until then, the 960,000-GPU figure is a design ceiling and the WPP result a vendor-reported early test.

FAQ

When can customers use Google Cloud A5X? Google says general availability arrives in November 2026. Ineffable Intelligence already has access.

Will Google deploy 960,000 Vera Rubin GPUs in November? No. Google says Virgo can scale to 80,000 GPUs at one site and 960,000 across sites. Those are architectural limits.

What is Confidential G4? A G4 VM with NVIDIA RTX PRO 6000 Blackwell GPUs that encrypts data and model weights in hardware. It is available in 36 zones.

References

• Google Cloud and NVIDIA expand co-engineering collaboration to scale infrastructure for the agentic era — Google Cloud (Oct. 21, 2026)
• Google Cloud and NVIDIA Announce General Availability of NVIDIA Vera Rubin Infrastructure — NVIDIA Blog (Oct. 21, 2026)
• Confidential G4 GPUs are GA — Google Cloud (Oct. 21, 2026)
• Google Cloud Sets Nov Availability for Vera Rubin A5X — Converge Digest (Oct. 25, 2026)