Crusoe

What Crusoe does

Crusoe is an AI factory company focused on accelerating the abundance of energy and intelligence. They provide reliable, scalable, cost-effective, and energy-first solutions for AI infrastructure. This involves harnessing large-scale energy resources, building AI-optimized data centers, and delivering an AI cloud platform. Crusoe aims to empower customers to build the future faster by offering high-performance NVIDIA & AMD compute, accelerated storage, optimized RDMA networking, and managed services for AI workloads. They are vertically integrated, from finding innovative energy sources to building and managing hyperscale AI factories and offering a scalable AI cloud platform. Crusoe prioritizes scale and speed alongside renewable and low-carbon energy use, pioneering sustainable AI infrastructure.

Products

• Crusoe CloudA purpose-built AI cloud platform offering GPU compute, orchestration, and managed services for large-scale AI training and inference workloads. It features high-performance NVIDIA & AMD GPUs, accelerated storage, optimized networking, and managed Kubernetes/Slurm.
• Crusoe Intelligence FoundryA platform for fine-tuning and deploying state-of-the-art models, offering serverless fine-tuning, one-click inference deployment, and a curated selection of top-performing models. It provides managed inference for AI applications with no infrastructure overhead.
• Serverless Fine-TuningA feature within Crusoe Intelligence Foundry that allows users to quickly adapt models with proprietary data, offering a serverless approach from data upload to an improved model in a few clicks, without cluster provisioning or surprise bills.
• Serverless InferenceA feature for serving models with ultra-low latency and high throughput, offering self-serve deployments and tailored deployments for greater flexibility. It integrates leading LLMs and generative models into applications with pay-as-you-go pricing.
• Crusoe Managed Kubernetes (CMK)A fully managed Kubernetes cluster that simplifies deployment and scaling of AI applications across GPU and CPU resources, offering rapid cluster provisioning and topology-aware orchestration.
• Crusoe Managed Slurm (CMS)Managed Slurm clusters pre-configured with Slurm workload manager, NVIDIA drivers, and job scheduler for immediate job submission, supporting migration of existing HPC workloads.
• Crusoe AutoClustersA feature that automates node recovery and job resumption for fault-tolerant scaling, continuously monitoring cluster health and replacing nodes with hot spares.
• Crusoe Command CenterA unified operations platform for AI workloads, providing real-time cluster health visibility across the full fleet and enabling governance and management of training jobs.
• Crusoe Edge ZoneDelivers high-performance AI Infrastructure, powered by Crusoe Spark modular AI data centers.
• Crusoe Spark modular AI data centersModular AI data centers powered by the world's largest second-life EV battery system, used for Crusoe Edge Zones.
• Crusoe MemoryAlloy™ technologyProprietary technology powering Crusoe Managed Inference, featuring a cluster-wide KV cache that eliminates redundant context recomputation for low latency and high throughput.
• GPU instancesAccess to high-performance GPUs including NVIDIA GB200 NVL72, NVIDIA B200 HGX, NVIDIA H200, NVIDIA H100, NVIDIA A100 (SXM & PCIe), NVIDIA L40S, AMD MI355X, and AMD MI300X, with various pricing options.
• CPU instancesGeneral-purpose and storage-optimized CPU instances for data processing, model checkpointing, and orchestrating GPU clusters.
• StorageReliable, low-latency storage options including Persistent Disks, Shared Disks, Object Storage (S3-compatible), and Ephemeral Disks, designed for massive datasets and high-throughput AI workloads.
• Self-Serve DeploymentsAllows users to spin up dedicated endpoints for open and fine-tuned models in minutes without sales engagement.
• Tailored DeploymentsDedicated, benchmarked endpoints optimized for specific models and workloads, with SLA-backed availability, optimized throughput, and enterprise observability.
• Provisioned ThroughputEnsures guaranteed throughput for generative AI applications, transacted via AI Model Units (AMUs).

Key capabilities

• Energy-first approach to AI infrastructure
• AI-optimized data centers
• Scalable AI cloud platform
• High-performance NVIDIA & AMD GPUs (GB200 NVL72, B200 HGX, H200, H100, A100, L40S, MI355X, MI300X)
• Accelerated storage
• Optimized RDMA networking
• Serverless Fine-Tuning
• Serverless Inference
• Managed Kubernetes (CMK)
• Managed Slurm (CMS)
• Fault-tolerant AutoClusters
• Crusoe Command Center for unified operations
• Proprietary MemoryAlloy™ technology for inference
• Low latency, high throughput (up to 9.9x faster TTFT, 5x higher throughput vs vLLM)
• Zero ingress and egress fees for data processing
• Petabyte-scale NFS backed by VAST Data
• GPU-accelerated instances for data engineering
• AI-first network backbone with hybrid full-mesh iBGP and MPLS-TE
• SOC 2 Type II, ISO 27001, and ISO 42001 certified compliance
• 99.5% uptime and resilient infrastructure
• 24/7 enterprise-grade support
• Flexible consumption models (pay-as-you-go, spot, on-demand, reserved pricing)
• One-click API generation for inference
• Intelligent routing and KV caching
• Automatic scaling for inference and training workloads
• Support for PyTorch, Kubeflow, and Ray
• NVIDIA NGC™-ready and AMD ROCm™ software environments
• S3-compatible object storage
• High-performance block storage
• Closed-loop, non-evaporative cooling systems for data centers
• High efficiency plumbing fixtures in data centers
• Intensive leak detection and monitoring in data centers
• Landscaping with native plants and no permanent irrigation for data centers

Use cases

• Building and deploying AI models faster
• Experimentation to production scale AI
• Customizing top models with proprietary data
• Deploying models up to 20x faster
• Cutting costs by up to 81%
• Eliminating operational overhead for AI infrastructure
• Data processing for AI at scale
• Streamlining data ingestion from external databases, APIs, object stores
• Automating data quality checks and sanitization
• Executing heavy data transformations
• Validating and publishing curated datasets
• Accelerating data engineering
• Training large-scale AI models
• Orchestrating GPU clusters
• Running open models
• Serving proprietary models on managed inference infrastructure
• Integrating LLMs and generative models into applications
• Developing and deploying AI agents
• Predicting global weather patterns (Jua)
• Generating personalized skincare routines (Perbangga Duha)
• Diagnosing production incidents with AI SRE agents (Alexander Sorrell)
• Scaling infrastructure confidently while maintaining healthy unit economics (Windsurf)
• Real-time model serving to hundreds of thousands of users (Odyssey)
• Building and deploying far more powerful and responsive AI agents (Roey Lalazar)

AI approach

Crusoe provides an energy-first AI factory company offering AI infrastructure, cloud platform, and managed AI services. They enable customers to train, fine-tune, and deploy state-of-the-art models, including open-source and proprietary ones, on high-performance NVIDIA and AMD GPUs. Their proprietary MemoryAlloy™ technology optimizes inference performance.

Tech named: NVIDIA GB200 NVL72, NVIDIA B200 HGX, NVIDIA H200, NVIDIA H100, NVIDIA A100 SXM, NVIDIA A100 PCIe, NVIDIA L40S, AMD MI355X, AMD MI300X, Crusoe Intelligence Foundry, Crusoe Cloud, Crusoe Managed Kubernetes, Crusoe Managed Slurm, Crusoe AutoClusters, Crusoe Command Center, MemoryAlloy™ technology, Apache NiFi, Airbyte, rclone, Great Expectations, ydata-profiling, Apache Spark, KubeRay, Kafka, VAST Data, NVIDIA drivers, Quantum-2 InfiniBand networking, AMD GPU support via ROCm, PyTorch, Kubeflow, Ray, NVIDIA NGC™, Terraform, S3-compatible object storage, InfiniBand, Infinity Fabric, vLLM

Industries served

• Technology, Information and Internet
• AI
• Cloud
• Data Centers
• Energy
• Infrastructure
• Manufacturing
• Environment
• Sustainability

What it says sets it apart

• Energy-first approach: vertically integrated from energy sourcing to cloud platform, prioritizing renewable and low-carbon energy use.
• Proprietary MemoryAlloy™ technology for ultra-low latency and high-throughput inference.
• Cost-effective performance: AI-optimized hardware and lightweight virtualization cut waste and unlock performance.
• Scalability: designed for massive demands of AI, with fault-tolerant scaling and automatic resource management.
• Simplified operations: Managed Kubernetes, Managed Slurm, and AutoClusters eliminate operational overhead.
• Reliability: 99.5% uptime, resilient infrastructure, 24/7 enterprise-grade support, 100% customer satisfaction.
• Comprehensive data processing solutions with zero ingress/egress fees and petabyte-scale NFS.
• Compliance: SOC 2 Type II, ISO 27001, and ISO 42001 certified.
• Focus on sustainability: minimizes water usage in data centers, invests in local communities, and supports new energy technologies.
• Access to latest GPU offerings: NVIDIA GB200 NVL72, HGX B200, AMD MI355x, MI300x.
• Flexible pricing models: pay-as-you-go, spot, on-demand, reserved pricing, and provisioned throughput.

Funding rounds we track

From the AI funding tracker — rounds as reported by the linked publications.