# Crusoe

Official pricing: https://www.crusoe.ai/cloud/pricing  
Category: gpu-cloud · Isolation: vm

## Pricing regimes (raw)

- **H100 SXM, on-demand (×8 node)** (resource): $0/vCPU-h, $0/GiB-h
- **H100 SXM, spot (×8 node)** (resource): $0/vCPU-h, $0/GiB-h [spot, sales]
- **H200 SXM, on-demand (×8 node)** (resource): $0/vCPU-h, $0/GiB-h
- **H200 SXM, spot (×8 node)** (resource): $0/vCPU-h, $0/GiB-h [spot, sales]
- **A100-80GB SXM, on-demand (×8 node)** (resource): $0/vCPU-h, $0/GiB-h
- **A100-80GB SXM, spot (×8 node)** (resource): $0/vCPU-h, $0/GiB-h [spot, sales]
- **MI300X, on-demand (×8 node)** (resource): $0/vCPU-h, $0/GiB-h
- **MI300X, spot (×8 node)** (resource): $0/vCPU-h, $0/GiB-h [spot, sales]

## Features

Yes: snapshots, fork/clone, pause/resume, persistent disk, volumes, ≥24 h sessions, custom image (Docker or snapshot), start from your own snapshot, full VM (own kernel), Docker inside, root, SSH, public IPv4, raw TCP inbound, egress allowlist, open internet, static egress IP, private networking, SOC 2, HIPAA, SSO, GPU, arm64, Python, audit logs, EU data residency, MCP server, snapshots on demand, gVisor or VM (no shared kernel), inbound access rules, firewall inside (nftables), extra volumes, shared volumes

No: memory snapshots, idle auto-stop, your own Docker/OCI image, nested virtualization, browser, desktop GUI, computer-use API, browser + desktop control, anti-bot stealth, CAPTCHA solving, residential IPs, preinstalled agents, HTTPS preview URLs, self-hosting, BYOC, open source, Windows, macOS, Node.js, HTTP method/path egress rules, wake on request, live resize, live fork (no pause), memory fork, spend limits, automatic snapshots, agent harness API, hosted agent API (their own agent), secret proxy, secret proxy for any API

Unknown: everything else. Evidence (source + quote) per feature: https://battleships.dev/data/providers/crusoe.json → feature_evidence

## Caveats

- GPU VM bundle includes matching CPU/RAM; billing per second in running state including lifecycle scripts. Stopped VM compute free, retained 128GiB OS disk $0.08/GiB-month until deletion. Reserved commitment bills in every state. Spot quote via sales. No network transfer charges currently.
- Any null count, CPU/RAM, billing increment or minimum is unverified, not unlimited/free.
- This is GPU-product research, not a claim that every feature of the entire vendor documentation was audited. Unverified features are null.
- GPU multi-count bundles, variant/region selection and fractional slices require a SKU-aware estimator; unsupported rows remain unpriced in strict cards.
- Modes with gpu=null are intentionally excluded, not free GPU modes. Never substitute 0 for a null rate. Flags alt indicate DIY/ML infrastructure, not the same thing as managed untrusted-code sandboxes.
- Unknown fee, storage or bandwidth means full monthly total may be unavailable. Rates exclude taxes.
- 2026-09-28: H100/H200/A100 SXM VMs exist only as 8x types (docs.crusoecloud.com: h100-80gb-sxm-ib.8x 176 vCPU / 960 GB); priced per GPU with gpu_min_count 8. Multi-GPU PCIe A100/L40S VMs priced with gpu_min_count = GPUs per VM. alt removed from all GPU VMs.
- Every VM has a fixed 128 GiB OS persistent disk ($0.08/GiB-month = $10.24/VM-month) billed while the VM exists, running or stopped; OS disks cannot be resized. Custom images billed at $0.08/GiB-month of used data. Shared disks $0.07/GiB-month (min 1 TiB). Object storage $0.06/GiB-month. Container registry $0.10/GiB-month. Managed Kubernetes $0.10/cluster-hour.
- GPU and s1a/s2a local NVMe (ephemeral) disks are erased when the VM is stopped and restarted; only the OS disk and persistent/shared disks survive a stop.
- Introductory $5 free credit applies only to Managed AI (Serverless Inference / Fine-Tuning), not to VMs. Enabling billing for VMs requires a valid non-prepaid credit card.
- Reservations (sales) bill for the whole term in every state; usage above the reserved GPU count at any instant bills on-demand; unused capacity does not roll over. Enterprise support tier is free only with qualifying annual commit/reservation.

## How this provider charges

As of 2026-09-28. GPU virtual machines and reserved capacity.

GPU VM bundle includes matching CPU/RAM; billing per second in running state including lifecycle scripts. Stopped VM compute free, retained 128GiB OS disk $0.08/GiB-month until deletion. Reserved commitment bills in every state. Spot quote via sales. No network transfer charges currently.

| Regime | When applicable | Billing | Numbers ($/GPU-hour unless slice) | Source |
|---|---|---|---|---|
| on-demand  | H100 SXM; count [8]; us-east1-a/us-southcentral1-a/eu-iceland1-a | per-second; min unknowns | 3.9; list; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| spot  | H100 SXM; count [8]; us-east1-a/us-southcentral1-a/eu-iceland1-a | per-second; min unknowns | unknown; sales-unpublished; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | H200 SXM; count [8]; eu-iceland1-a | per-second; min unknowns | 4.29; list; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| spot  | H200 SXM; count [8]; eu-iceland1-a | per-second; min unknowns | unknown; sales-unpublished; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | A100-80GB SXM; count [8]; us-east1-a | per-second; min unknowns | 2.3; list; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| spot  | A100-80GB SXM; count [8]; us-east1-a | per-second; min unknowns | unknown; sales-unpublished; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | MI300X ; count [8]; us-east1-a | per-second; min unknowns | 3.45; list; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| spot  | MI300X ; count [8]; us-east1-a | per-second; min unknowns | unknown; sales-unpublished; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | MI355X ; count [8]; us-east2-a | per-second; min unknowns | unknown; sales-unpublished; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| spot  | MI355X ; count [8]; us-east2-a | per-second; min unknowns | unknown; sales-unpublished; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | B200 SXM; count [8]; eu-iceland1-a/eu-norway1-a/us-west1-a | per-second; min unknowns | unknown; sales-unpublished; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| spot  | B200 SXM; count [8]; eu-iceland1-a/eu-norway1-a/us-west1-a | per-second; min unknowns | unknown; sales-unpublished; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | B300 SXM; count [8]; eu-iceland1-a | per-second; min unknowns | unknown; sales-unpublished; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| spot  | B300 SXM; count [8]; eu-iceland1-a | per-second; min unknowns | unknown; sales-unpublished; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | GB200 NVL; count [4]; eu-iceland1-a | per-second; min unknowns | unknown; sales-unpublished; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| spot  | GB200 NVL; count [4]; eu-iceland1-a | per-second; min unknowns | unknown; sales-unpublished; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | A100-80GB PCIe; count [1]; us-east1-a; L40S also us-southcentral1-a | per-second; min unknowns | 2; list; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | A100-80GB PCIe; count [2]; us-east1-a; L40S also us-southcentral1-a | per-second; min unknowns | 2; list; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | A100-80GB PCIe; count [4]; us-east1-a; L40S also us-southcentral1-a | per-second; min unknowns | 2; list; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | A100-80GB PCIe; count [8]; us-east1-a; L40S also us-southcentral1-a | per-second; min unknowns | 2; list; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | L40S PCIe; count [1]; us-east1-a; L40S also us-southcentral1-a | per-second; min unknowns | 1.5; list; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | L40S PCIe; count [2]; us-east1-a; L40S also us-southcentral1-a | per-second; min unknowns | 1.5; list; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | L40S PCIe; count [4]; us-east1-a; L40S also us-southcentral1-a | per-second; min unknowns | 1.5; list; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | L40S PCIe; count [8]; us-east1-a; L40S also us-southcentral1-a | per-second; min unknowns | 1.5; list; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |
| on-demand  | L40S PCIe; count [10]; us-east1-a; L40S also us-southcentral1-a | per-second; min unknowns | 1.5; list; per_gpu_hour | [source](https://www.crusoe.ai/cloud/pricing) |

## Gotchas
- GPU VM bundle includes matching CPU/RAM; billing per second in running state including lifecycle scripts. Stopped VM compute free, retained 128GiB OS disk $0.08/GiB-month until deletion. Reserved commitment bills in every state. Spot quote via sales. No network transfer charges currently.
- Any null count, CPU/RAM, billing increment or minimum is unverified, not unlimited/free.
- This is GPU-product research, not a claim that every feature of the entire vendor documentation was audited. Unverified features are null.
- GPU multi-count bundles, variant/region selection and fractional slices require a SKU-aware estimator; unsupported rows remain unpriced in strict cards.

## Worked example (required protocol workload)
4 vCPU / 8 GiB, 50 concurrent × 8 hours/day × 22 days = **8,800 instance-hours/month**. 30% CPU utilization does not reduce allocated GPU uptime. 50 GiB snapshots and 100 GiB egress are additional. The requested CPU-only workload has no GPU type/count, so its full GPU-provider total is **null**, not a fictitious CPU equivalent.

Explicit GPU sensitivity example: add **1 × A100-80GB** to each of those 50 workers at the A100-80GB quoted shape. Compute component = 8,800 × $2 = **$17,600.0000**. CPU/RAM are included in that SKU (not independently resizable).
This is compute-only, before platform fees, storage, egress, setup, minimum rounding, cold-start/idle tails and capacity quotas. 50 concurrent GPUs are not promised by a unit rate.

## Evidence scope
Sources read are listed in the provider/features JSON. Raw pricing pages, relevant docs and API payloads are retained under `../raw/`. All unknown feature toggles remain null.