Compute

Virtual machines and block storage for any AI and ML workloads.

Flexible capacity planning

Get access to the latest NVIDIA GPU platforms or CPU-only servers and balance reserved, and on-demand pricing models that align perfectly with your needs.

AI performance without penalty

Receive bare-metal-level performance from dedicated hosts — we do not virtualize and share GPU and network cards.

InfiniBand-powered AI clusters

Create a multi-host clusters for AI workloads with non-blocking NVIDIA Quantum InfiniBand fabric. It delivers 3.2 Tbit/s throughput per 8-GPU host and ensures direct GPU-to-GPU communication.

AI-ready operational system

Save time when creating instances or configuring a cluster for AI workloads by using an AI/ML-ready image that contains pre-installed GPU and network drivers, to start a GPU-accelerated environment quickly.

Network storage volumes

Reduce the cluster recovery time by leveraging network disks mounted to every virtual instance. This provides cloud-native elasticity and a quick VM restart when failure occurs.

Integrated monitoring

Receive detailed information about your cluster and virtual machine performance by using our integrated Monitoring service. Our dashboards display AI-specific metrics alongside general system performance data.

GPU host configurations

Early access

NVIDIA GB200 NVL72

  • 72x GB200 GPU 384GB
  • 36x Grace CPU with 72 Arm® Neoverse™ V2 cores
  • Up to 17 TB LPDDR5X
  • 14.4 Tbit/s InfiniBand
  • Ubuntu 24.04 LTS for NVIDIA® GPUs (CUDA® 12)
Early access

NVIDIA B200

  • 1x or 8x B200 GPU 180GB SXM
  • 16x or 128x vCPU Intel Emerald Rapids
  • 224 or 1792 GB DDR5
  • 3.2 Tbit/s InfiniBand
  • Ubuntu 22.04 LTS for NVIDIA® GPUs (CUDA® 12)
Available

NVIDIA H200

  • 1x or 8x H200 GPU 141GB SXM
  • 16x or 128x vCPU Intel Sapphire Rapids
  • 200 or 1600 GB DDR5
  • 3.2 Tbit/s InfiniBand
  • Ubuntu 22.04 LTS for NVIDIA® GPUs (CUDA® 12)
Available

NVIDIA H100

  • 1x or 8x H100 GPU 80GB SXM
  • 16x or 128x vCPU Intel Sapphire Rapids
  • 200 or 1600 GB DDR5
  • 3.2 Tbit/s InfiniBand
  • Ubuntu 22.04 LTS for NVIDIA® GPUs (CUDA® 12)
Available

NVIDIA L40S (Intel)

  • 1x L40S GPU 48GB PCIe
  • 8x or 40x vCPU Intel Xeon Gold
  • 32 or 160 GB DDR5
  • Ubuntu 22.04 LTS for NVIDIA® GPUs (CUDA® 12)
Available

NVIDIA L40S (AMD)

  • 1x L40S GPU 48GB PCIe
  • 16x or 192x vCPU AMD EPYC
  • 96 or 1152 GB DDR5
  • Ubuntu 22.04 LTS for NVIDIA® GPUs (CUDA® 12)

CPU host configurations

Available

Intel

  • 2x or 48x vCPU Intel Xeon Gold 6338
  • 8 or 192 GB DDR5
  • Ubuntu 22.04 LTS
Available

AMD

  • 4x or 128x vCPU AMD EPYC 9654
  • 16 or 512 GB DDR5
  • Ubuntu 22.04 LTS

Try self-service console

Up to 32 NVIDIA GPUs are available immediately via web console

Block network storage

Choose one of three options of network disks that differ by performance, reliability and pricing:

  • SSDs with no data replication,
  • SSDs with erasure coding,
  • SSDs with data mirroring.

Observability and monitoring

Control the cluster state and detect performance issues early, by using our monitoring capabilities. We display a wide range of performance metrics, from GPU utilization to InfiniBand network parameters on the web UI dashboards or as pre-assembled Grafana dashboards.

Getting started

Create and manage GPU clusters on the cloud platform on your own or contact us to learn more about working with one of our experts.