Crusoe pricing: GPU rental rates
checked 21 November 2026
Crusoe rents 7 GPU models, from $1.50 an hour (L40S, on-demand). Its cheapest on-demand H100 is $3.90 (H100, on-demand), 14% above the $3.43 market median. Storage, data transfer and billing rules are under the prices.
- Cheapest GPU $1.50 L40S, on-demand
- H100 market median $3.43 Crusoe's H100 is 14% above it
- 7 offers
- 7 GPU models
- Source crusoe.ai
In our latest check: None of its 7 matching prices changed since 14 November 2026.
Crusoe Cloud is an AI infrastructure platform for GPU instances, CPU instances, storage, managed Kubernetes and managed inference. It fits teams that want NVIDIA and AMD datacenter GPUs, RDMA-backed networking, S3-compatible storage and the option to use on-demand, spot or reserved capacity. Public pricing is unusually explicit about storage, managed Kubernetes, ingress and egress treatment.
Compare the full daily GPU market
What else you pay at Crusoe
Besides the GPU rate: Storage costs $0.08/GiB/month persistent disk; $0.07/GiB/month shared disk. Data in and out is free. Billing is per second. A stopped instance still bills for its storage. No minimum rental time (no minimum commitment on demand). Each of these was checked on Crusoe's own documentation on 16 October 2026. Add the charges that apply to you in the job cost calculator.
Crusoe prices, checked 21 November 2026
More filters: provider, memory, node size, stock, sort, price unit
| GPU / provider | Billing | $/GPU-hour | Node · total per hour | Rent |
|---|---|---|---|---|
| L40S · Crusoe48 GB | on-demand | $1.50 | – | Rent → |
| A100 PCIe 80GB · Crusoe80 GB | on-demand | $2.00 | – | Rent → |
| A100 SXM 80GB · Crusoe80 GB | on-demand | $2.30 | – | Rent → |
| MI300X · Crusoe192 GB | on-demand | $3.45 | – | Rent → |
| H100 · Crusoe80 GB | on-demand | $3.90 | – | Rent → |
| H200 (SXM) · Crusoe141 GB | on-demand | $4.29 | – | Rent → |
| B300 · Crusoe288 GB | on-demand | $8.96−9.5% in 30 days | – | Rent → |
Try another provider or reduce the minimum memory.
Fresh prices come first. Within each billing type, prices run lowest first, with unavailable offers last. Commercial relationships never affect ordering. How we collect and rank prices.
Same GPU, provider, billing and price on several hosts or listings: one row, with the count. – means the provider does not state it.
Monthly view assumes 730 hours of continuous use. Daily view assumes 24 hours.
Cite this data
GPU Rental Prices, "Crusoe cloud GPU pricing: rates and total costs", prices checked 21 November 2026, https://gpurentalprices.com/providers/crusoe
Quote it with a link to this page. Downloads of the whole dataset are on Data & API.
Cheapest listed per-GPU-hour rates verified at the source. Before storage, egress and networking.
Crusoe access and company facts
| Check | Published information | Checked |
|---|---|---|
| Minimum account funding | not published | 16 October 2026 |
| Founded | 2018 | 17 October 2026 |
| Regions | Seven regions: Houston, Ohio, Sparks (Nevada) and Virginia in the United States, two in Iceland and one in Norway | 18 October 2026 |
| SOC 2 | Type II | 17 October 2026 |
| ISO 27001 | certified | 17 October 2026 |
| Data location terms | US, Iceland, Norway | 17 October 2026 |
A company fact or certification is not a guarantee for every product or region. Unrecorded fields are left unknown. The smallest recorded node does not establish immediate stock or account approval.
Strengths and trade-offs
Strengths
Computed from the verified fees and platform matrix (cells checked between 16 October 2026 and 17 October 2026 at Crusoe's own documentation; verified items link to their source). Full matrices: fees · enterprise.
Pricing model
Crusoe's pricing page lists reserved, on-demand and spot pricing for GPU instances, plus pay-as-you-go and provisioned-throughput options for managed inference. The page says GPU and CPU instances are billed by the minute, and reserved capacity is a custom agreement for committed resource volume and guaranteed availability. Newer or high-end GPU SKUs may require contacting sales.
Capacity tiers
Crusoe exposes GPU instances, CPU instances, storage, managed Kubernetes and managed inference. The Cloud overview describes on-demand, spot and reserved capacity options, from single-node experiments to multi-thousand GPU clusters. Crusoe Managed Kubernetes and AutoClusters are positioned for fault-tolerant orchestration across large GPU and CPU resource pools.
GPU lineup
Crusoe's current public pricing and product navigation list NVIDIA GB200 NVL72, B200 HGX, H200 HGX, H100 HGX, A100 SXM and PCIe, L40S, AMD MI355X and AMD MI300X. The overview page says the platform provides the latest NVIDIA and AMD GPUs optimized for AI training at multiple scales. Some SKUs are contact-sales rather than fully self-serve.
Storage, egress and hidden costs
Crusoe lists persistent disks, shared disks, container registry usage and object storage separately from compute. The pricing FAQ says Crusoe does not currently charge for network ingress or egress within a VPC or to and from the public internet. Managed Kubernetes has its own cluster pricing, and managed inference has separate token or throughput pricing, so total cost should include more than the GPU line item.
Reliability and support posture
Crusoe positions the platform for production AI, with enterprise support, proactive monitoring, isolated environments, high-performance networking and scalable storage. Its Cloud page gives public support-response and customer-satisfaction claims, but buyers should validate contractual support terms, covered services and any service-credit remedies in their own agreement. Public docs also expose API, Terraform, quota, audit-log and monitoring surfaces useful for production operations.
Free credits and referral programs
A current public free-credit, startup-credit or referral program for Crusoe Cloud was not found in the official public pages reviewed. Not publicly documented.
Who should pick Crusoe
Pick Crusoe if you need AI-optimized infrastructure with NVIDIA and AMD datacenter GPUs, RDMA networking, managed Kubernetes and a clear zero-ingress and zero-egress posture. It is strongest for teams planning production training or inference that can choose between on-demand, spot and reserved capacity. Confirm sales-led SKU availability and contract terms if you need guaranteed access to newer GPU families.
Similar providers
RunPod
Secure Cloud (datacenter, reliable) and Community Cloud (peer hosts, cheaper). Per-second billing, serverless option.
from $0.24/hr secure cloud
6 GPU families in common
Oracle Cloud
Oracle Cloud Infrastructure GPU shapes at list price, the same in every public region. Most come only as whole bare-metal servers of 4 or 8 GPUs; the A10 and V100 also rent one at a time.
from $2.00/hr on-demand
6 GPU families in common
Ranked by overlap in the GPU models both providers list today; not an endorsement and never affected by commercial relationships.
FAQ
Yes. Crusoe publishes spot as a GPU instance pricing option for supported SKUs.
Yes. Crusoe's pricing FAQ says GPU and CPU instances are billed by the minute.
Crusoe's pricing FAQ says it does not currently charge network ingress or egress within a VPC or to and from the public internet.
A current public credit or referral program was not found in official public pages. Not publicly documented.
Profile checked 6 July 2026. Sources: www.crusoe.ai · www.crusoe.ai · docs.crusoecloud.com · docs.crusoecloud.com · docs.crusoecloud.com