On-demand instances available hourly; 1-Click Clusters from $8.87 to $9.86 per GPU per hour for 2 weeks to 1 year; reserved capacity available at lower prices with annual commitment.
Crusoe offers flexible pricing with on-demand hourly billing (no minimum commitment), spot pricing for fault-tolerant workloads, and reserved capacity options. GPU instances range from $1.50 to $4.29 per GPU-hour depending on model. Serverless Inference uses per-token pricing (input, output, and cac
Key features
Superclusters with NVIDIA GB300 NVL72 and Quantum-2 InfiniBand
1-Click Clusters™ with NVIDIA HGX B200 and H100 GPUs
On-demand GPU instances
Single-tenant, shared-nothing architecture
SOC 2 Type II certification
Managed cluster orchestration
Co-engineering support
Crusoe Intelligence Foundry with serverless fine-tuning
Managed Kubernetes and Managed Slurm for simplified operations
99.5% uptime with 24/7 enterprise-grade support
Multiple GPU options including NVIDIA GB200, B200, H200, H100 and AMD MI355X, MI300X
Serverless Inference with up to 9.9x faster time to first token
Command Center unified operations platform
Closed-loop, non-evaporative cooling systems for water efficiency
What makes it different
Complete AI factories integrating power, cooling, and GPUs in one system designed for peak AI performance
Dedicated single-tenant infrastructure for training and inference at scale
Expert co-engineering from team building infrastructure for world's largest AI labs
Rack-scale systems optimized for agentic AI and reasoning workloads
Hardware-level isolation with caged clusters
Vertically-integrated from energy generation to cloud platform
Minimal water impact with closed-loop cooling systems
Purpose-built cloud infrastructure optimized for AI performance