Crusoe

What's good about Crusoe

• Great GPU availability for on-demand use
• Published hourly on-demand rates for most GPU types
• Renewable and low-carbon energy sourcing with vertical integration from power to cloud
• 99.5% uptime commitment with 24/7 enterprise support
• Managed inference and serverless fine-tuning available alongside raw GPU instances
• Fast instance deployment

Crusoe GPU prices

7 GPU models across 7 configurations. Rank compares Crusoe's cheapest on-demand listing for each GPU with every provider's cheapest.

What a run costs on Crusoe

24 hours at today's cheapest listing for each GPU, at the size it is listed.

• 1× H100 SXM for 24 hours

  At $3.90/GPU-hr

  $93.60

  $16.80 over the market median rate

• 1× H200 for 24 hours

  At $4.29/GPU-hr

  $102.96

  $5.16 under the market median rate

• 1× A100 SXM for 24 hours

  At $2.30/GPU-hr

  $55.20

  $16.92 over the market median rate

The market median is every provider's cheapest on-demand price for that GPU.

Crusoe LLM API prices

Pay-per-token pricing, per 1M tokens.

Prices last updated November 3, 2026

Billing & terms

On-Demand Instances

Pay by the hour for maximum agility and unthrottled compute

Spot Instances

Use spare capacity at significant discounts compared to on-demand pricing

Reserved Instances

Lock in guaranteed resources at lowest rates with custom commitments

Pay-as-you-go Inference

Flexible pricing for LLM inference based on token usage

Provisioned Throughput

Guaranteed inference capacity with AI Model Units, priced lower for longer commitments

Dedicated Inference Deployments

Hourly per-GPU pricing for self-serve dedicated endpoints, with monthly and volume rates via sales

Serverless Fine-Tuning

Per-token pricing tiered by the parameter count of the base model being tuned

Regions

Data centers sited near low-cost energy in the United States, including a large-scale campus in Abilene, Texas, plus European capacity in Iceland and Norway

Support

Documentation, learning hub, and 24/7 enterprise-grade support with a 99.5% uptime commitment; SOC 2, ISO 27001, and ISO 42001 certifications

Watch out for

• Contact sales required for newest GPU models (GB200, B200, MI355X)
• Spot capacity is quoted by sales rather than published as a self-service rate
• Data center footprint concentrated in the United States with limited European capacity

Features

Energy-First Infrastructure

Vertically integrated power sourcing across wind, solar, hydropower, geothermal, natural gas, and carbon capture

Fast Spin Up

Machines are deployed quickly when they are launched

Persistent Storage

Persistent and shared disks, object storage, and a container registry for AI datasets

SXM Support

Provides instances with NVIDIA SXM GPU interconnects

SLA Guarantee

99.5% uptime commitment backed by 24/7 enterprise-grade support

Managed Inference

LLM inference with up to 9.9x faster time-to-first-token and 5x higher throughput versus vLLM

Serverless Fine-Tuning

LoRA-based supervised fine-tuning of open models through Crusoe Intelligence Foundry, with no cluster provisioning

Cluster Orchestration

Managed Kubernetes, Managed Slurm, and fault-tolerant AutoClusters for multi-node training

Command Center

Unified operations platform for high-performance AI workloads

Security Certifications

ISO 27001 and ISO 42001 certified for information security and responsible AI governance

Services

• GPU Instances

  High-performance GPU instances powered by NVIDIA GPUs

• CPU Instances

  General-purpose and storage-optimized CPU instances for data processing

• Managed Inference

  Fully managed LLM inference spanning serverless, self-serve dedicated, and tailored deployments

  • Open models including DeepSeek V4, GLM 5.2, Kimi K2.6, Llama 3.3, Gemma 4, Qwen3, GPT-OSS, and Nemotron 3
  • Serverless Inference billed per million input, output, and cached tokens
  • Self-Serve Deployments give dedicated H100 or H200 endpoints without sales engagement
  • Tailored Deployments are benchmarked and optimized with Crusoe's team
  • Provisioned throughput with AI Model Units (AMUs)
  • Up to 9.9x faster time-to-first-token and 5x higher throughput versus vLLM

• Serverless Fine-Tuning

  LoRA-based supervised fine-tuning of open models in Crusoe Intelligence Foundry

  • No cluster provisioning required
  • Token-based pricing tiered by base model parameter count
  • One-click deployment of the tuned model to an inference endpoint
  • Full portability of the resulting weights

• Managed Kubernetes and Slurm

  Managed Kubernetes (CMK) and Managed Slurm clusters, plus fault-tolerant AutoClusters, for AI workloads

  • Simplifies deployment and scaling
  • Supports both GPU and CPU resources
  • Billed per cluster hour on top of instance costs
  • Container registry integration

• Reserved Clusters

  Large-scale reserved GPU clusters for enterprise and research workloads

  • Custom node configurations
  • Flexible interconnect options
  • Long-term reservation for consistent availability

• Crusoe Edge Zones

  High-performance AI infrastructure powered by Crusoe Spark modular AI data centers

  • Scalable edge computing capabilities
  • Modular AI data center deployment
  • Reduced latency for edge AI workloads

Getting started

1. Step 1

   Create an account

   Sign up for Crusoe Cloud services through their website

2. Step 2

   Select GPU instance

   Choose from available GPU types including L40S, A100, H100, H200, B200, GB200, and AMD options

3. Step 3

   Configure storage

   Set up persistent storage options for your workloads

4. Step 4

   Launch instance

   Deploy via the console, or provision programmatically with the Crusoe CLI or Terraform