NVIDIA partners expand European cloud capacity and agent controls

NVIDIA cloud partners are expanding AI infrastructure across Europe, the Nordics, Africa and Central Asia, giving developers more options to keep workloads close to their data. The investments support model training, customization and inference, alongside robotics applications and controls over what AI agents can access.

Nscale plans to deploy NVIDIA Vera Rubin infrastructure across five European markets. In tests, NVIDIA DSX MaxLPS enabled 37% more GPUs and nearly 50% higher aggregate inference throughput within the same provisioned power budget. Crusoe is developing liquid-cooled modular facilities for Vera Rubin, while Mistral plans to receive the infrastructure in 2027.

Providers are also expanding services built around open models. Nebius is integrating inference hardware into its Token Factory and supporting robotics development through its Physical AI Workbench. Baseten developed a specialized code-review routing model for CodeRabbit, improving routing agreement from 75.8% to 80.7% while reducing estimated inference costs by 50% for the evaluated workload. Fireworks and Era4 are partnering to serve enterprises across the U.K. and Europe.

For agent security, NVIDIA introduced the OpenShell Reference Architecture at GTC Berlin. It combines secure runtime software, inference systems, CPUs for agent execution and hardware-based enforcement to restrict access and actions. Partners including CoreWeave, DigitalOcean, Nebius and Nscale support the architecture, which is designed to help organizations deploy agents within defined operational boundaries.