tools
CoreWeave Cloud
CoreWeave Cloud provides GPU and CPU compute, storage, networking, and tools for training, deploying, and operating AI workloads.
In inglese
CoreWeave Cloud is an AI-focused cloud platform for AI labs, software companies, enterprises, and research teams. It provides on-demand and spot GPU and CPU instances, object and file storage, networking, orchestration, inference, observability, and security features.
Customers use it to train and serve models, run simulations, build AI agents, and manage production workloads. Pricing is usage-based rather than subscription-based; some capacity and services require contacting CoreWeave, and dedicated inference is listed as private preview in the documentation.
Features
- On-demand and spot NVIDIA GPU instances
- On-demand and spot AMD CPU instances
- AI Object Storage with Hot, Warm, Cold, and Archive tiers
- Distributed File Storage for AI workloads
- VPC, NAT Gateway, public IP, and Direct Connect networking
- Mission Control for orchestration and workload management
- Dedicated Inference with OpenAI-compatible requests
- REST, gRPC, S3-compatible, and AI Object Storage APIs
Use cases
- Train large language and other foundation models
- Serve models through dedicated or managed inference
- Run reinforcement learning and agent evaluations
- Store datasets, model weights, checkpoints, and logs
- Run high-performance computing and engineering simulations
- Move workloads across clouds and regions
Pros
Cons
Latest updates
- New Observability
Mission Control Agent is now available in public preview; Mission Control MCP is now generally available.
- Update Platform
GB200 NVL72-powered instances, GB300 NVL72 Quantum-X InfiniBand, and GB300 NVL72 Spectrum-X RoCE: New default 595 → 580.
- Update CKS
CKS supports v1.37; new clusters use v1.37; v1.34 deprecated; Events retained 30 minutes; Cilium LB IPAM allocates IP addresses.
- New Billing
The FOCUS Billing Export API is now available in public preview. Pull hourly CoreWeave usage in FOCUS 1.2 JSON or CSV format.
- Update Platform
New default: 580 → 595 for GB300 NVL72 Quantum-X InfiniBand, GB300 NVL72 Spectrum-X RoCE, and RTX PRO 6000 Blackwell Server Edition.
Capabilities
- API — “This is only possible because of CoreWeave’s Kubernetes API and range of high-performance compute and storage options.” source
- Runs models for you — “Run low-latency AI inference with high-performance compute” source
- Builds agents and workflows — “CoreWeave helps teams build, evaluate, deploy, and monitor models and agents—speeding time to production with experiment tracking, governance and guardrails, and workflow integrations.” source
- Traces and evaluates — “From bare metal to token, CoreWeave provides the constant observability essential to fast progress—plus built-in agent development and AIOps tools that evaluate, monitor, and iterate agentic AI systems during development and production.” source
Get it
Security
- GDPR — “Yes. CoreWeave’s privacy and security program is designed to support customer compliance with applicable data protection regulations, including GDPR and CCPA, where relevant.” source
- Data kept in the EU — “CoreWeave operates data centers across multiple geographic regions, including the United States and Europe, to support customer requirements for performance, availability, and data residency.” source
Pricing
- Prices checked
- 2026-09-24
NVIDIA GB300 NVL72
Price on request
- On-demand GPU instances
NVIDIA GB200 NVL72
- $42.00 / Hour (North America On-Demand)
- $10.50 / Hour (North America Inference Single CPU Price)
- $42.00 / Hour (Europe On-Demand)
- On-demand GPU instances
NVIDIA HGX B300
- $35.84 / Hour (North America Spot)
- $36.70 / Hour (Europe Spot)
- On-demand GPU instances
NVIDIA HGX B200
- $68.80 / Hour (North America On-Demand)
- $34.11 / Hour (North America Spot)
- $8.60 / Hour (North America Inference Single CPU Price)
- $68.80 / Hour (Europe On-Demand)
- $34.87 / Hour (Europe Spot)
- On-demand GPU instances
NVIDIA RTX PRO 6000 Blackwell Server Edition (High Memory)
- $20.00 / Hour (North America On-Demand)
- $11.09 / Hour (North America Spot)
- $2.50 / Hour (North America Inference Single CPU Price)
- $20.00 / Hour (Europe On-Demand)
- $11.01 / Hour (Europe Spot)
- On-demand GPU instances
NVIDIA RTX PRO 6000 Blackwell Server Edition (Standard Memor
- $9.56 / Hour (North America Spot)
- $9.56 / Hour (Europe Spot)
- On-demand GPU instances
NVIDIA GH200
- $6.50 / Hour (North America On-Demand)
- $6.50 / Hour (North America Inference Single CPU Price)
- $6.50 / Hour (Europe On-Demand)
- On-demand GPU instances
NVIDIA HGX H100
- $49.24 / Hour (North America On-Demand)
- $19.71 / Hour (North America Spot)
- $6.16 / Hour (North America Inference Single CPU Price)
- $49.24 / Hour (Europe On-Demand)
- $19.51 / Hour (Europe Spot)
- On-demand GPU instances
NVIDIA L40
- $10.00 / Hour (North America On-Demand)
- $6.27 / Hour (North America Spot)
- $1.25 / Hour (North America Inference Single CPU Price)
- $10.00 / Hour (Europe On-Demand)
- On-demand GPU instances
NVIDIA HGX H200
- $50.44 / Hour (North America On-Demand)
- $20.93 / Hour (North America Spot)
- $6.31 / Hour (North America Inference Single CPU Price)
- $50.44 / Hour (Europe On-Demand)
- $20.64 / Hour (Europe Spot)
- On-demand GPU instances