Bare metal & clusters

Whole servers.
Whole clusters.

When one GPU isn't the unit any more: single-tenant 8-GPU servers with no virtualization layer, and multi-node clusters for distributed training. Quoted by an engineer, not a sales team.

Typical nodeHGX B200
GPUs8 × NVIDIA B200
GPU memory8 × 180 GB HBM3e
GPU interconnectNVLink 5
Between nodesInfiniBand

Exact specs and regions confirmed in your quote

Options

Two ways
to go bigger.

Both run under the same account as your hourly GPUs.

Bare-metal servers

A physical 8-GPU server assigned to you. No hypervisor, no neighbours, full NVLink bandwidth between its GPUs.

  • Order from the console and pay by bank transfer
  • Monthly terms
  • Root on the host

GPU clusters

Multiple nodes on an InfiniBand fabric for distributed training and large-scale inference, sized with our engineers.

  • From a few nodes to hundreds of GPUs
  • Terms from months to years
  • Topology and storage planned with you
Get a quote

Tell us the shape
of the job.

An engineer reads every request and replies with availability, regions, networking and price, usually within a day.

Prefer email? Write to help@packet.ai.

Your next run
starts at $0.39.

Sign up, top up, launch. You'll be on a GPU in minutes.

No contractHourly or flat monthlyTalk to a human any time