NVIDIA said its cloud partner ecosystem across Europe, the Nordics, Africa and Central Asia is adding regional AI capacity for training, inference and agent deployment, as developers seek compute closer to their data and more control over where workloads run. The company’s latest update also highlights new infrastructure, model-serving and security layers meant to support so-called sovereign AI clouds and agentic applications.
Sovereign capacity and efficiency claims
In the source material published by NVIDIA on 21 October 2026, the company said partners including Nscale, Crusoe, Aleria, Cassava Technologies, Firebird, G42, Radiant, SWI Compute and Vultr are expanding NVIDIA-powered AI infrastructure across the region. NVIDIA said Nscale is using the NVIDIA DSX Platform to coordinate power allocation with workload requirements and, in tests using NVIDIA GB300 NVL72 systems, demonstrated that NVIDIA DSX MaxLPS enabled 37% more GPUs and nearly 50% higher aggregate inference throughput within the same provisioned power budget. NVIDIA also said Nscale plans to deploy NVIDIA Vera Rubin infrastructure for frontier-model training and production workloads across five European markets.
NVIDIA said the broader regional capacity combines the DSX AI Factory platform with Vera Rubin, BlueField and Spectrum-X Ethernet networking, with the stated aim of giving customers more choice over where workloads and data reside. The company framed the expansion as part of Europe’s effort to turn local energy, data, research and industrial capability into AI infrastructure.
Open models, inference and production use cases
On the open-model side, NVIDIA said AI clouds are expanding infrastructure and services for developers building with open models and libraries. The company said Nebius has adopted NVIDIA Groq 3 LPX and is integrating it with Vera Rubin NVL72 on Nebius Token Factory using Spectrum-X Ethernet networking. According to NVIDIA, that setup is intended to support faster token generation for latency-sensitive applications and agentic systems through managed serverless and dedicated endpoints.
NVIDIA also said Nebius is offering a Physical AI Workbench for robotics and physical AI, with companies including Rhoda AI, RoboForce and Sereact building on the platform. The source material further said Nebius has launched three NVIDIA retail agent blueprints covering inventory, store operations and fine-tuning into early access on its infrastructure, and is expanding services through a Physical AI Living Lab and in Germany as a computer supplier for SPRIND’s Next Frontier Challenge.
Elsewhere, NVIDIA said Baseten developed a specialized model for code-review routing using Nemotron 3.5 Lightning, which CodeRabbit used to improve routing agreement from 75.8% to 80.7% while reducing estimated inference costs by 50% for the evaluated workload. The company also said Fireworks and Era4 are working together to bring open models to enterprises in the U.K. and Europe, with liquid-cooled capacity powered by NVIDIA Blackwell Ultra GPUs at Era4’s Chesterfield facility.
NVIDIA said Mistral is expanding its accelerated computing infrastructure to support model training, customization and inference, including planned Vera Rubin infrastructure for delivery in 2027. The company said Mistral Large 4 was trained on 3,800 NVIDIA Grace Blackwell GPUs in Mistral-operated European data centers. NVIDIA also said Verda plans to deploy Vera Rubin while promoting open weight models, with Nemotron and Cosmos already in production and the Dynamo open-source framework used for inference optimization.
Agent controls and deployment boundaries
NVIDIA said the partner ecosystem is also supporting the OpenShell Reference Architecture, announced at GTC Berlin, which is designed to give organizations full-stack control over AI agents and clear boundaries around what they can access and do. The company said the architecture is built on OpenShell secure runtime software and uses Vera Rubin NVL72 for model inference, Vera CPUs with OpenShell for agent harnesses and tool use, and BlueField-4 DPUs running NVIDIA Sentry for in-silicon enforcement.
Partners named by NVIDIA as supporting the OpenShell Reference Architecture include Aleria, Baseten, Cassava, CoreWeave, Crusoe, DigitalOcean, Firebird, Fireworks, Firmus, G42, GMI Cloud, IREN, Lambda, Lightning AI, Nebius, Nscale, Radiant Verda and Vultr. NVIDIA said the framework is intended to help customers deploy secure agentic systems with defined access controls as models are connected to tools and business systems.
The company’s update comes alongside a separate 21 October 2026 NVIDIA Newsroom announcement on DSX POD AI factories for Europe, which extends the same infrastructure theme into pre-integrated modular containers. The DSX POD note said its 14-day deployment claim excludes site preparation and described early-2027 deliveries as a forward-looking plan.
Be the first to comment