#1
Vast.aiMarketplace
Configuration8× H100 · 192 vCPU
RegionIndia
PurchaseOn-demand
$1.56/ GPU / hr
NVIDIA · Hopper
H10080 GB HBM3
A widely deployed accelerator for large-model training and high-throughput inference.
Lowest observed rate$1.56per GPU / hour · sample feedView H100 on Flatkey Compute ↗
ArchitectureHopper
VRAM80 GB
Memory bandwidth3.35 TB/s
Published2022
Price board
5 configurations tracked
Median sample rate: $2.59 per GPU-hour.
#2
HyperstackGPU cloud
Configuration8× H100 · 252 vCPU
RegionCanada
PurchaseOn-demand
$2.50/ GPU / hr
#3
RunpodGPU cloud
Configuration8× H100 · 64 vCPU
RegionGlobal
PurchaseOn-demand
$2.59/ GPU / hr
#4
AWSHyperscaler
Configuration8× H100 · 192 vCPU
RegionUS East
PurchaseSpot
$3.78/ GPU / hr
#5
LambdaGPU cloud
Configuration1× H100 · 26 vCPU
RegionUnited States
PurchaseOn-demand
$4.29/ GPU / hr