Crusoe Deepens NVIDIA Alliance to Build Next-Gen AI Infrastructure and Inference Stack

New updates have been reported about Crusoe.

Crusoe is expanding its strategic collaboration with NVIDIA to position itself as a fully integrated, NVIDIA-native AI infrastructure provider for the emerging agentic AI market, aligning its entire stack with NVIDIA’s hardware and model roadmap to boost scalability and performance for enterprise customers. As part of this roadmap, Crusoe will be an early adopter of the NVIDIA Vera CPU, deploying it alongside NVIDIA Rubin GPUs and NVIDIA Vera Rubin systems in Crusoe Cloud beginning in late 2026 and continuing through 2027 to keep post-training agentic AI workloads fully utilized at large scale.

Crusoe will also use NVIDIA’s Omniverse DSX Blueprint and the Vera Rubin DSX Reference Design to plan and operate its next wave of gigawatt-scale AI factories, applying digital twins, AI-driven power and cooling optimization, and advanced mechanical and electrical design to deliver energy-optimized, redundancy-focused infrastructure. On the model side, Crusoe is integrating NVIDIA Nemotron 3 Super and Nemotron 3 VoiceChat into its Intelligence Foundry and Crusoe Managed Inference service, leveraging its MemoryAlloy-based platform and a high-performance Rust tokenizer integrated with NVIDIA’s open-source Dynamo library to materially cut time-to-first-token and increase throughput for long-context, multi-agent, and voice-interactive applications, while contributing these performance gains back to the open-source ecosystem.