Overview
Crusoe Cloud provides a high-performance, sustainable cloud platform with the latest hardware.
Virtual Machine (VM) specifications
GPU-enabled VMs
Customers can create GPU-enabled VMs with the following specs:
| Type | vCPU | GPU | Memory | Ephemeral Disk | VPC Network | InfiniBand Network | Block Storage Network | Zones |
|---|---|---|---|---|---|---|---|---|
a100-80gb.1x | 12 vCPUs Intel Xeon (Ice Lake) | 1x NVIDIA A100 80GB PCIe | 120GB | 1x 960GB NVMe | 21.875 Gbps | N/A | 3.125 Gbps | us-west1-a, us-east1-a |
a100-80gb.2x | 24 vCPUs Intel Xeon (Ice Lake) | 2x NVIDIA A100 80GB PCIe | 240GB | 2x 960GB NVMe | 43.75 Gbps | N/A | 6.25 Gbps | us-west1-a, us-east1-a |
a100-80gb.4x | 48 vCPUs Intel Xeon (Ice Lake) | 4x NVIDIA A100 80GB PCIe | 480GB | 4x 960GB NVMe | 87.5 Gbps | N/A | 12.5 Gbps | us-west1-a, us-east1-a |
a100-80gb.8x | 96 vCPUs Intel Xeon (Ice Lake) | 8x NVIDIA A100 80GB PCIe | 960GB | 8x 960GB NVMe | 175 Gbps | N/A | 25 Gbps | us-west1-a, us-east1-a |
a100-80gb-sxm-ib.8x | 96 vCPUs Intel Xeon (Ice Lake) | 8x NVIDIA A100 80GB SXM4 | 960GB | 8x 960GB NVMe | 175 Gbps | 1600 Gbps | 25 Gbps | us-east1-a |
h100-80gb-sxm-ib.8x | 176 vCPUs Intel Xeon (Sapphire Rapids) | 8x NVIDIA H100 80GB SXM5 | 960GB | 8x 960GB NVMe | 175 Gbps | 3200 Gbps | 25 Gbps | us-east1-a, us-southcentral1-a, eu-iceland1-a |
h200-141gb-sxm-ib.8x | 176 vCPUs Intel Xeon (Sapphire Rapids) | 8x NVIDIA H200 141GB SXM5 | 2000GB | 8x 1.92TB NVMe | 175 Gbps | 3200 Gbps | 25 Gbps | eu-iceland1-a |
l40s-48gb.1x | 8 vCPUs AMD EPYC (Genoa) | 1x NVIDIA L40S 48GB PCIe | 147GB | N/A | 17.5 Gbps | N/A | 2.5 Gbps | us-east1-a, us-southcentral1-a |
l40s-48gb.2x | 16 vCPUs AMD EPYC (Genoa) | 2x NVIDIA L40S 48GB PCIe | 294GB | N/A | 35 Gbps | N/A | 5 Gbps | us-east1-a, us-southcentral1-a |
l40s-48gb.4x | 32 vCPUs AMD EPYC (Genoa) | 4x NVIDIA L40S 48GB PCIe | 588GB | N/A | 70 Gbps | N/A | 10 Gbps | us-east1-a, us-southcentral1-a |
l40s-48gb.8x | 64 vCPUs AMD EPYC (Genoa) | 8x NVIDIA L40S 48GB PCIe | 1176GB | N/A | 140 Gbps | N/A | 20 Gbps | us-east1-a, us-southcentral1-a |
l40s-48gb.10x | 80 vCPUs AMD EPYC (Genoa) | 10x NVIDIA L40S 48GB PCIe | 1470GB | N/A | 175 Gbps | N/A | 25 Gbps | us-east1-a, us-southcentral1-a |
mi300x-192gb-ib.8x | 240 vCPUs AMD EPYC (Genoa) | 8x AMD MI300X 192GB OAM | 2000GB | 8x 1.92TB NVMe | 175 Gbps | 3200 Gbps | 25 Gbps | us-east1-a |
b200-180gb-sxm-ib-8x | 176 vCPUs Intel Xeon (Emerald Rapids) | 8x Nvidia B200 180GB SXM | 3000GB | 8x 1.92TB NVMe | 175 Gbps | 3200 Gbps | 25 Gbps | eu-iceland1-a |
gb200-186gb-nvl-4x | 128 vCPUs Nvidia ARM Neoverse (Grace) | 4x Nvidia GB200 186GB NVL | 886GB | 4x 1.92TB NVMe | 175 Gbps | 1600 Gbps | 25 Gbps | eu-iceland1-a |
CPU VMs
Crusoe Cloud offers two CPU-only instance types available in all locations: c1a and s1a. c1a instances are "general purpose" and intended to be used for control planes, web serving, and other lightweight compute. s1a instances are "storage optimized" and intended for running high performance file systems like Lustre or ceph, or object stores like minio.
Customers can create c1a VMs with the following specs:
| Type | vCPU | Memory | Disk | VPC Network | Block Storage Network |
|---|---|---|---|---|---|
c1a.2x | 2 vCPUs AMD (Genoa) | 8GB | N/A | 1 Gbps | 512 Mbps |
c1a.4x | 4 vCPUs AMD (Genoa) | 16GB | N/A | 2 Gbps | 1 Gbps |
c1a.8x | 8 vCPUs AMD (Genoa) | 32GB | N/A | 5 Gbps | 2 Gbps |
c1a.16x | 16 vCPUs AMD (Genoa) | 64GB | N/A | 10 Gbps | 4 Gbps |
c1a.32x | 32 vCPUs AMD (Genoa) | 128GB | N/A | 20 Gbps | 10 Gbps |
c1a.64x | 64 vCPUs AMD (Genoa) | 256GB | N/A | 35 Gbps | 20 Gbps |
c1a.128x | 128 vCPUs AMD (Genoa) | 512GB | N/A | 70 Gbps | 40 Gbps |
c1a.176x | 176 vCPUs AMD (Genoa) | 704GB | N/A | 100 Gbps | 50 Gbps |
Customers can create s1a VMs with the following specs:
| Type | vCPU | Memory | Disk | VPC Network | Block Storage Network |
|---|---|---|---|---|---|
s1a.20x | 20 vCPUs AMD (Genoa) | 176GB | 1x 12.8TB NVMe | 25 Gbps | up to 12.5 Gbps |
s1a.40x | 40 vCPUs AMD (Genoa) | 352GB | 2x 12.8TB NVMe | 50 Gbps | up to 25 Gbps |
s1a.60x | 60 vCPUs AMD (Genoa) | 528GB | 3x 12.8TB NVMe | 75 Gbps | up to 37.5 Gbps |
s1a.80x | 80 vCPUs AMD (Genoa) | 704GB | 4x 12.8TB NVMe | 100 Gbps | up to 50 Gbps |
s1a.120x | 120 vCPUs AMD (Genoa) | 1056GB | 6x 12.8TB NVMe | 150 Gbps | up to 75 Gbps |
s1a.160x | 160 vCPUs AMD (Genoa) | 1408GB | 8x 12.8TB NVMe | 200 Gbps | up to 100 Gbps |
Certain VM types in certain regions are currently restricted and may not be available for immediate provisioning. If you require access, please contact our sales team to discuss your use case.
All local storage associated with GPU and s1a instances are considered ephemeral storage, even when configured in a storage cluster. Please see managing ephemeral disks for more details.
A vCPU represents a single thread on a CPU.
VM lifecycle
VMs transition through the following states:
| State | Description |
|---|---|
Creating | The VM is being created. This can take up to a minute. After the VM has been created, it will transition to the Stopped state |
Starting | The VM is being started. Once it has started, it will transition to Running. |
Running | The VM is running and accessible. You can now SSH into it and run your desired software. The VM will continue to run until you stop it. |
Stopping | The VM is being shut down, and no more customer work will be performed. Once it has shut down, it will transition to Stopped. |
Stopped | The VM is stopped and inaccessible. You must manually start it to change the state. Stopped VMs aren't billed for hourly compute charges. |
VM billing
Crusoe Cloud bills for VM usage in two separate ways:
- On-demand
- Reservations
On-demand usage provides developers the most flexibility. VMs are charged based on the time the machine is in the running state (per second). This includes time that the VM is spent running Lifecycle Scripts on startup or shutdown. Stopped VMs are not charged for on-demand usage, but you will still be billed for the 128GB OS disk at $0.08/GiB/month until the VM is deleted.
Reservations offer discounts in exchange for committed usage. These VMs are billed for the entire duration of the commitment, regardless of their state. Learn more about reservations.
A VM may only be billed via one of these mechanisms (on-demand or reservation) at a time, but different VMs in the same organization may be billed differently.