NVIDIA announced that its cloud partners are significantly expanding AI computing capabilities across Europe, Northern Europe, Africa, and Central Asia to support AI model training, inference, and the development of secure AI agents. Nscale, leveraging the NVIDIA DSX platform, increased GPU count by 37% and inference throughput by nearly 50% in tests. Crusoe is developing liquid-cooled modular AI factories for the NVIDIA Vera Rubin platform. Nebius is combining NVIDIA Groq 3 LPX and Vera Rubin NVL72 to accelerate token generation. Mistral is expanding its NVIDIA accelerated computing infrastructure, with plans to deliver NVIDIA Vera Rubin infrastructure in 2027, and has already used 3,800 NVIDIA Grace Blackwell GPUs to train the Mistral Large 4 model. NVIDIA also introduced the NVIDIA OpenShell reference architecture to ensure the security of AI agents.