CIComputeRadar
Live multi-GPU configuration index

8× H200 GPU Server Rental Prices

Compare 8× NVIDIA H200 GPU systems for large-model training and inference.

Available offers3
Providers2
Lowest listed price now$16.00 / hour
Last catalog refresh2026-09-07 21:50 UTC

At the latest catalog refresh, the lowest listed matching offer is $16.00 / hour from Verda. Availability and final checkout pricing can change after this timestamp.

Vultr4181Hivelocity2695UpCloud2610OVHcloud2081Gcore1816Scaleway864Sesterce719Alibaba Cloud549Cherry Servers381Servers.com291Exoscale234Spheron200Vast.ai150Ori142Verda128Timeweb Cloud100Hyperstack79Northflank57Akash Network43RunPod42bunny.net Storage30Prime Intellect21IDrive e219Replicate19Baseten18SaladCloud17Wasabi16Paperspace15Render15Thunder Compute13Modal11Hostwinds10Melbicom10Fly.io8fal.ai7Worldstream7Impossible Cloud6Limestone Networks6Serverspace6GMI Cloud5Together AI5InterServer4Railway4TensorDock4Hyperbolic3Filebase1Storj1
⌕
3 matching offers17,643 indexed · 47/47 sources with data · catalog 12:50 AM · showing 1–3
VerdaGPUSpot

H200 SXM5 141GB

Location not specified · Catalog listed · live capacity not verified anonymously

$16.00/ hour≈ $11,680.00/mo
CPU176 vCPU176 physical cores
RAM1,360 GBsystem memory
GPU8× H200141 GB VRAM / GPU
StorageSee detailsdynamic
Network—not reported
Available · updated 12:49 AM
Pricing noteFull specs
VerdaGPUOn-demand

H200 SXM5 141GB

Location not specified · Catalog listed · live capacity not verified anonymously

$32.00/ hour≈ $23,360.00/mo
CPU176 vCPU176 physical cores
RAM1,360 GBsystem memory
GPU8× H200141 GB VRAM / GPU
StorageSee detailsdynamic
Network—not reported
Available · updated 12:49 AM
Pricing noteFull specs
ReplicateGPUcommitted-spend capacity

Replicate · gpu-h200-8x

Location not specified · Available

$43.92/ hour≈ $32,061.60/mo
CPU—CPU not reported
RAM—system memory not reported
GPU8× H200141 GB VRAM / GPU
StorageManagedReplicate publishes hardware CPU/RAM/GPU specs but not a fixed disk capacity per hardware SKU
NetworkManaged Replicate networking · fixed per-hardware port speed/traffic allowance is not publishedprovider-reported network terms; port speed not reported
Available · updated 12:43 AM
Pricing noteFull specs
How to read this page

Compare the full configuration, not only the headline price

This page requires at least 8 NVIDIA H200 GPUs in the normalized offer. It is designed for searches where the complete multi-GPU configuration matters, not only a per-GPU rate.

GPU count does not prove NVLink, NVSwitch, InfiniBand or equivalent topology. Verify interconnect, host RAM, local storage and networking before ordering.

Selection notes

What to check before renting

01

Check whether the displayed price is per GPU or for the full host.

02

Confirm NVLink/NVSwitch or other interconnect requirements with the provider.

03

For training clusters, compare CPU, RAM, storage throughput and network fabric as well as GPU count.

Related searches

Compare nearby options

GPU configuration8× H100 GPU servers→GPU configuration4× H100 GPU servers→GPU configuration2× H100 GPU servers→GPU configuration8× B200 GPU servers→GPU configuration8× A100 GPU servers→
FAQ

Common questions

Where can I rent 8× H200 GPU servers?

The live matching offers above come from currently connected provider sources and are refreshed with the catalog.

Does an 8-GPU result guarantee all GPUs are on one host?

It represents an offer reporting at least the requested GPU count, but provider topology can vary. Confirm host and interconnect details at checkout.