1
Thunder Compute
$0.35/GPU-hr
RTX A600048GB VRAM
AmpereRegion unavailable
Verify pricing on the provider's official site before checkout.
Loading GPU prices…
GPU cloud price radar
Find the cheapest GPU cloud for your AI workload.
Compare H100, A100, L40S, and RTX 4090 listings by hourly price, region, and workload fit — built for indie hackers and small AI teams.
- 0
- GPU models tracked
- 0
- Tracked listings
- $0.00
- Lowest $/GPU-hr
Fast answer
Cheapest option by GPU
RTX A6000
Region unavailableThunder Compute · Thunder RTX A6000 (1x GPU, undefined)
$0.35/GPU-hr→
RTX 4090
KRiwinv · iwinv 4090.G1 (1x GPU, VAT excl.)
$0.51/GPU-hr→
A100
JPGPUSOROBAN · GPUSOROBAN t40-1-a A100 40GB (1x GPU)
$0.63/GPU-hr→
L4
USAWS · g6.xlarge (1x GPU)
$0.81/GPU-hr→
L40S
Region unavailableDataCrunch · Verda L40S 48GB (1x GPU)
$1.37/GPU-hr→
Refreshed Jul 27, 2026, 7:56 PM UTC
Vast.ai GPU marketplace
Explore GPUs beyond the tracked list
Browse the Vast.ai marketplace directly, then confirm the current price, host reliability, and rental terms before starting an instance.
Explore Vast.ai
External link. ComputeHunt may receive a reward when you use it.
Popular GPUs
Lowest price on trending GPUs
RTX A6000
Thunder Compute
$0.35/GPU-hr
$0.35/GPU-hr
RTX 4090
iwinv
$0.51/GPU-hr
$0.51/GPU-hr
A100
GPUSOROBAN
$0.63/GPU-hr
$0.63/GPU-hr
L4
AWS
$0.81/GPU-hr
$0.81/GPU-hr
L40S
DataCrunch
$1.37/GPU-hr
$1.37/GPU-hr
Showing 5 of 11
Curated in-demand models (H200, B300, RTX PRO 6000, etc.) — one cheapest listing each. Bar length is relative within this set.
Cost decision
GPU rental vs managed API
See when renting a GPU beats per-token or per-image API pricing at your monthly volume.
Open cost calculator →GPU rental$61.60
Managed API$320.00
Illustrative sample at 8h/day, 22 days a month. Enter your own numbers in the calculator.
Live rankings
Cheapest tracked GPUs
Ranked by verified GPU-hour price and freshness from tracked provider data.
Updated Jul 27, 2026, 7:56 PM UTC
2
iwinv
$0.51/GPU-hr
RTX 409024GB VRAM
Ada LovelaceKR
Verify pricing on the provider's official site before checkout.
3
GPUSOROBAN
$0.63/GPU-hr
A10080GB VRAM
AmpereJP
Verify pricing on the provider's official site before checkout.
4
AWS
$0.81/GPU-hr
L424GB VRAM
Ada LovelaceUS
Verify pricing on the provider's official site before checkout.
Workload fit
Pick your workload
Jump straight to recommendations for inference, image generation, video, or batch jobs.
Movement
Recent price drops
Hourly price decreases detected between tracked snapshots.
How it works
Built for trustworthy price decisions
ComputeHunt ranks by tracked price data and keeps source checks visible before you rent.
Price decides the order
Listings are sorted by verified GPU-hour price, then freshness. When a provider reports stock, in-stock listings rank higher.
Fresh, timestamped data
The footer shows the latest data refresh for the page. Price drops compare the latest snapshot against the previous one, so you see real movement.
Official source checks
Every listing keeps a provider source and sends users to the provider site for final pricing, availability, and terms.
Answers
GPU cloud cost FAQ
Quick answers to the questions AI builders ask before renting compute.
What is the cheapest H100 cloud GPU right now?+
It changes constantly. ComputeHunt ranks tracked H100 listings by verified GPU-hour price and freshness, with stock as a tie-breaker when providers report it. The cheapest option from the current tracked data appears at the top.
See H100 comparison →What is the difference between marketplace and fixed-price GPU rental?+
Marketplace pricing can move sharply by host, stock, and region, while fixed-price providers are usually more predictable. Compare price, stock, and region on each GPU page before choosing for your workload.
Compare providers →What is the cheapest RTX 4090 cloud for ComfyUI?+
The RTX 4090 (24GB) is a popular value pick for ComfyUI and SDXL. Check the RTX 4090 page for ranked listings, or the image-generation workload page for VRAM-aware recommendations.
Image generation picks →When is renting a GPU cheaper than a managed API?+
Roughly when your monthly volume is high and steady enough that fixed GPU hours beat per-unit API pricing. Enter your hours and volume in the calculator to see the exact breakeven point.
Open the calculator →Which GPU cloud is best for Stable Diffusion?+
For SD and SDXL, prioritize 16GB+ VRAM and a low hourly rate — RTX 4090 and L40S are common picks. The image-generation workload page ranks current listings that meet the VRAM minimum.
Best for Stable Diffusion →What does the listed price mean?+
The main price is normalized to one GPU for one hour. Multi-GPU instances show their available configs separately so you can compare both per-GPU price and total instance cost. USD is canonical; Korean and Japanese views show converted amounts.
Are these prices live?+
Prices come from enabled provider APIs and source-labeled tracked listings. The page footer shows the latest data refresh time, and provider checkout pages remain the final source of truth.