Source-linked
Every sample offer carries a provider or pricing-page link for follow-up verification.
GPU market intelligenceLabeled launch sample
The price of compute,
made comparable.
Compare sample GPU rental prices per hour across providers, regions, and purchase models. Review the listed configuration context before you shortlist a cloud.
MARKET / ON-DEMAND SAMPLE
Lowest sample rateH100
$1.56/ GPU / HR
A100 80GB$0.88floor
L40S$0.69floor
Configurations17sample feed
H10080 GBH200141 GBB200180 GBA100 80GB80 GBL40S48 GBRTX 409024 GB
Market scanner / sample feed
Find capacity without opening thirty tabs.
Sample rates · observed Oct 12, 2026
| GPU / provider | Configuration | Region | Rental | Availability | Per GPU / hr |
|---|---|---|---|---|---|
| RTX 4090Vast.ai | 1× · 12 vCPU · 64 GB | EUEurope | On-demand | Available | $0.32$0.32 total |
| RTX 4090Runpod | 1× · 16 vCPU · 62 GB | GLGlobal | On-demand | Available | $0.59$0.59 total |
| L40SHyperstack | 1× · 16 vCPU · 60 GB | CACanada | On-demand | Available | $0.69$0.69 total |
| A100 80GBVast.ai | 4× · 64 vCPU · 512 GB | USUnited States | On-demand | Available | $0.88$3.52 total |
| A100 80GBRunpod | 1× · 16 vCPU · 125 GB | GLGlobal | On-demand | Available | $1.19$1.19 total |
| A100 80GBHyperstack | 1× · 28 vCPU · 180 GB | CACanada | On-demand | Available | $1.35$1.35 total |
| H100Vast.ai | 8× · 192 vCPU · 2268 GB | INIndia | On-demand | Limited | $1.56$12.48 total |
| H100Hyperstack | 8× · 252 vCPU · 1440 GB | CACanada | On-demand | Available | $2.50$20.00 total |
| H100Runpod | 8× · 64 vCPU · 512 GB | GLGlobal | On-demand | Limited | $2.59$20.72 total |
| H200Hyperstack | 8× · 252 vCPU · 1800 GB | CACanada | On-demand | Available | $3.42$27.36 total |
| H200Runpod | 8× · 128 vCPU · 1440 GB | GLGlobal | On-demand | Limited | $3.59$28.72 total |
| H100Lambda | 1× · 26 vCPU · 200 GB | USUnited States | On-demand | Limited | $4.29$4.29 total |
| B200Runpod | 8× · 180 vCPU · 1600 GB | GLGlobal | On-demand | Limited | $5.89$47.12 total |
Prices are normalized to USD per GPU-hour.How comparison works →
Ready to deploy the shortlist?Continue to Flatkey Compute to review current GPU options.
Continue to Flatkey Compute ↗GPU catalogue
Start with the silicon.
Compare the market around a specific accelerator—memory, architecture, pricing floor, and which clouds list it.
View all GPU rental prices →01Datacenter
H100
Hopper · 80 GB HBM3
From $1.56/hr5 offers ↗
02Datacenter
H200
Hopper · 141 GB HBM3e
From $3.42/hr3 offers ↗
03Datacenter
B200
Blackwell · 180 GB HBM3e
From $5.50/hr2 offers ↗
04Datacenter
A100 80GB
Ampere · 80 GB HBM2e
From $0.88/hr3 offers ↗
05Datacenter
L40S
Ada Lovelace · 48 GB GDDR6
From $0.69/hr2 offers ↗
06Consumer
RTX 4090
Ada Lovelace · 24 GB GDDR6X
From $0.32/hr2 offers ↗
Provider index
Know who owns the capacity.
Marketplace, specialist cloud, or hyperscaler—the operating model matters almost as much as the sticker price.
Compare all providers →01Vast.aiMarketplaceUnited States3 configurationsfrom $0.32↗02HyperstackGPU cloudUnited Kingdom4 configurationsfrom $0.69↗03RunpodGPU cloudUnited States6 configurationsfrom $0.59↗04LambdaGPU cloudUnited States1 configurationfrom $4.29↗05NebiusGPU cloudNetherlands2 configurationsfrom $3.85↗06AWSHyperscalerUnited States1 configurationfrom $3.78↗
GPU cloud regions
Location changes the real cost.
Compare regional price signals without losing the latency, data movement, and availability questions behind them.
Explore every region →Buyer field guide
The hourly rate is only one line in the bill.
Learn how to normalize per-GPU and node rates, separate spot from on-demand pricing, model egress and idle time, and request a quote procurement can verify.
Read the pricing guide →Built for scrutiny
A cheap number is useless without context.
Normalized
Sample rates are shown in USD per GPU-hour while preserving total-node cost and purchase type.
Status-labeled
Availability labels remain visible, but launch samples are never presented as live inventory.