At NVIDIA GTC Berlin on 22 October 2026, Google Cloud and NVIDIA announced an expanded collaboration on AI infrastructure, from the largest GPU clusters to confidential computing and on-premises AI.
What's new
- A5X instances on NVIDIA Vera Rubin NVL72: bare-metal GPU instances, with general availability planned for 6 November 2026 and wider availability in Europe, the Middle East and Africa for data residency.
- G4 Confidential instances: NVIDIA RTX PRO 6000 Blackwell GPUs with confidential computing, generally available.
- Sovereign AI on premises: NVIDIA HGX B300 systems on Google Distributed Cloud are generally available, with enterprise search with Gemini in preview for air-gapped environments.
Make AI do real work
We build AI that reads documents, answers questions from your own material and handles repetitive work. Start with a free readiness review.
Why it matters for your business
Few small or mid-sized businesses will rent a 72-GPU rack. What reaches them is the effect: more GPU capacity in more regions, more choice over where data is processed, and confidential options for sensitive workloads. If you train or run your own models, the practical questions are capacity, region and price per job, and those change fast.
How D3V can help
- We help you choose between managed AI services and your own GPUs, and size the infrastructure to the workload (Google Cloud infrastructure).
- Running AI on Kubernetes? We design, migrate and run GKE clusters (GKE managed services).
- Our free optimization assessment shows what your current setup really costs.
Source: Google Cloud infrastructure at NVIDIA GTC Berlin 2026, Google Cloud blog.
