Crusoe
Crusoe provides an AI cloud platform with GPU compute, managed inference, and serverless fine-tuning for training and serving models.
At a Glance
Engagement
Available On
Alternatives
Listed Oct 2026
About Crusoe
Crusoe describes itself as "the AI factory company," offering a vertically integrated AI infrastructure stack that spans a GPU cloud, managed AI services, data center infrastructure, and an energy-first approach to power. Its developer-facing products are Crusoe Cloud and Crusoe Intelligence Foundry, aimed at teams that train, fine-tune, and serve models.
What It Is
Crusoe Cloud is an AI cloud platform that provides NVIDIA and AMD GPU compute, accelerated storage, and RDMA networking. It includes Crusoe Managed Kubernetes, Crusoe Managed Slurm, and fault-tolerant Crusoe AutoClusters to reduce operational overhead. A Command Center offers a unified operations view for high-performance AI workloads, and Crusoe Edge Zones deliver infrastructure from modular data centers.
Crusoe Intelligence Foundry
Foundry is the managed AI layer. It offers a model hub with open models from several labs, including DeepSeek, GLM, Gemma, GPT-OSS, Llama, NVIDIA Nemotron, and Qwen, and teams can bring their own models. Serverless Fine-Tuning, announced as generally available, takes a user from data upload to an improved model without provisioning clusters. For inference, Crusoe offers Serverless Inference, Self-Serve Deployments, and Tailored Deployments, and says it delivers up to 9.9x faster time to first token and 5x higher throughput than vLLM.
Hardware and Developer Resources
The GPU lineup lists NVIDIA GB200 NVL72, B200, H200, and H100, plus AMD MI355X and MI300X. Developers get documentation, how-to guides, support, a cookbook, and demos on GitHub and Hugging Face, such as PyTorch distributed training on Slurm and Ray inference on Crusoe Managed Kubernetes.
Current Status
The site announces a $3.9B Series F and Crusoe's ISO 27001 and ISO 42001 certifications. Crusoe also states 99.5% uptime for Crusoe Cloud.
Community Discussions
Be the first to start a conversation about Crusoe
Share your experience with Crusoe, ask questions, or help others learn from your insights.
Pricing
GPU Instances (On-demand)
Hourly on-demand GPU pricing; spot and reserved options via sales. Representative: NVIDIA H100 80GB HGX at $3.90/GPU-hr.
- NVIDIA H200 141GB HGX: $4.29/GPU-hr
- NVIDIA H100 80GB HGX: $3.90/GPU-hr
- NVIDIA A100 80GB SXM: $2.30/GPU-hr
- NVIDIA A100 80GB PCIe: $2.00/GPU-hr
- NVIDIA L40S 48GB: $1.50/GPU-hr
- AMD MI300X 192GB: $3.45/GPU-hr
- GB200 NVL72, B200 HGX, AMD MI355X: Contact sales
CPU Instances
On-demand CPU instances priced per vCPU-hour.
- General-purpose: $0.04/vCPU-hr
- Storage-optimized: $0.09/vCPU-hr
Serverless Inference
Pay-as-you-go managed inference priced per 1M tokens (representative: DeepSeek V4 Pro input $1.74).
- DeepSeek V4 Pro: $1.74 input / $3.48 output / $0.15 cached per 1M tokens
- DeepSeek V4 Flash: $0.14 input / $0.28 output / $0.03 cached per 1M tokens
- GLM 5.3: $1.40 input / $4.40 output / $0.26 cached per 1M tokens
- GPT-OSS 120B: $0.05 input / $0.20 output / $0.05 cached per 1M tokens
- Nemotron 3 Ultra 550B: $1.00 input / $3.20 output / $0.25 cached per 1M tokens
Serverless Fine-Tuning
Fine-tuning priced per 1M tokens by model.
- Llama 3.1 8B Instruct: $0.40 per 1M tokens
- GPT-OSS 120B: $6.00 per 1M tokens
- DeepSeek V4 Flash: $5.00 per 1M tokens
- GLM 5.2 FP8: $10.00 per 1M tokens
Self-Serve Deployments
Dedicated endpoints priced per GPU hour.
- NVIDIA H100 80GB: $5.50/hr
- NVIDIA H200 141GB: $6.00/hr
- NVIDIA B200 180GB: $9.65/hr
- AMD MI355x 288GB: $7.40/hr
Contact sales
Reserved GPU capacity, Provisioned Throughput and Tailored Deployments via sales.
- Provisioned throughput transacted via AI Model Units (AMUs)
- Longer commitment lowers cost
Capabilities
Key Features
- NVIDIA and AMD GPU cloud compute
- Crusoe Intelligence Foundry managed inference
- Serverless Fine-Tuning
- Serverless Inference, Self-Serve and Tailored Deployments
- Crusoe Managed Kubernetes
- Crusoe Managed Slurm
- Crusoe AutoClusters
- Command Center operations platform
- Crusoe Edge Zones modular infrastructure
- Bring-your-own model support