Hardware

GPU servers, spec’d and built for your racks.

  • NVIDIA H100, H200, B200 and B300.
  • NVIDIA RTX PRO 5000, 5500 and 6000.
  • AMD Instinct MI300X, MI325X and MI355X.
  • From single parts to complete nodes.
  • Itemized quotes. Lead times quoted per build.
A 4U GPU server with its GPU tray pulled halfway out, showing eight H100 SXM GPUs, NVMe bays, a fan wall and front-panel InfiniBand ports.NVMe bays, 30 TBFront fan wall400G InfiniBand8× H100 80GB SXMNVLink switches

01What we ship

Parts, or a complete node.

  • Burn-in tested before it ships.
  • Matched to your workload and board.
  • Ready to rack and boot.
  1. 01

    GPUs

    NVIDIA H100, H200, B200, B300 and RTX PRO 5000, 5500 and 6000, and AMD Instinct MI300X, MI325X and MI355X, burn-in tested before they ship. Other GPUs quoted on request.

  2. 02

    GPU servers

    2U and 4U servers built around the GPUs, with chassis, CPUs, power and cooling matched. Ready to rack and boot.

  3. 03

    Memory

    High-density ECC DDR5 and DDR4, matched to your board and channel layout.

  4. 04

    Storage

    NVMe and enterprise SSDs and HDDs for scratch space, datasets and checkpoints, from a few drives to a full storage tier.

  5. 05

    Networking

    Switches, NICs, transceivers and cabling for InfiniBand or Ethernet fabrics across multi-node clusters.

  6. 06

    Full node builds

    A complete machine, spec’d, built and tested, then shipped ready to join your cluster or connect to HZ.

Sample configurationSpec’d to your workload
PartNVIDIA sampleAMD sample
GPUs8× H100 80GB SXM8× MI300X 192GB OAM
GPU fabricNVLink and NVSwitchInfinity Fabric
CPUs2× EPYC 90042× EPYC 9004
Memory1.5 TB DDR5 ECC2 TB DDR5 ECC
Storage30 TB NVMe30 TB NVMe
Network400G InfiniBand400G Ethernet (RoCE)
Chassis4U8U

02Process

From spec to racked.

Three steps. Help connecting to HZ once it's racked.

  1. 01

    Tell us what you need

    Share your workload, rough budget and timeline in the form below, or book a call.

  2. 02

    We spec and quote

    We design a build for your workload and send an itemized quote.

  3. 03

    Ship and connect

    The build ships to your datacenter. To connect it to HZ, we'll help you install the agent and bring it online.

03Support

Support after it ships.

  • Engineers on setup, drivers and tuning.
  • Or hand us the whole configuration.
  1. 01

    Support

    Work with our engineers on setup, drivers and tuning, or hand us the whole configuration.

  2. 02

    Lead times

    Lead times depend on configuration and stock.

04Quote

Tell us what you're building.

  1. Share your workload and budget.
  2. We spec the build and send back an itemized quote.

Prefer email? sales@hz-ai.io

Quote request
Opens your email app with these details filled in.