Crusoe Cloud

Crusoe Cloud is the AI infrastructure platform from Crusoe, a company founded in 2018 by Chase Lochmiller and Cully Cavness. It rents NVIDIA GPUs, including the H100, H200, B200 and B300, and AMD MI300X and MI355X accelerators. Multi-node clusters use InfiniBand networking and can be orchestrated with managed Kubernetes (CMK) or Slurm. It also offers general-purpose and storage-optimized CPU VMs, persistent and shared disks, object storage, a container registry, and managed AI services: serverless inference, self-serve dedicated endpoints and serverless fine-tuning. Crusoe says its platform runs on clean energy. It suits AI labs, startups training or serving models, and enterprises that want GPU capacity outside the hyperscalers. Pricing is usage-based. GPUs and CPUs are billed by the hour, storage per GiB-month, Kubernetes per cluster-hour and inference per million tokens, while reserved capacity is quoted by sales. The pricing page shows several self-serve GPU rates rising from December 1, 2026. Developers can work through a CLI, a Terraform provider, a REST API and an MCP server. SemiAnalysis gave Crusoe a Gold rating in its ClusterMAX system and praised the console and support. The main limitation is breadth. Crusoe has fewer managed services and a narrower geographic footprint than the hyperscalers, few independent user reviews, and service terms that reportedly exclude HIPAA workloads.