Skip to main content

About GPU Tracker

The Kayak for GPU Cloud Computing — real-time price comparison across major providers and marketplaces.

What is GPU Tracker?

GPU Tracker aggregates real-time GPU cloud pricing from the providers and marketplaces we track into a single searchable, filterable database. Compare H100 spot prices, find the best $/VRAM ratio, or discover which cloud has the cheapest A100 instances.

Data refreshes every 6 hours via GitHub Actions using the open-source gpuhunt library. All pricing is served as a public JSON endpoint — no tracking, no ads, no paywall.

How It Works

1
Provider APIs

Queried via gpuhunt + custom scrapers — AWS, Azure, GCP, OCI, Lambda Labs, RunPod, Vast.ai, Vultr, Cudo, Nebius, TensorDock, DigitalOcean, HotAisle, CloudRift, Verda, Scaleway, Hyperstack, Genesis Cloud, OVHcloud, CoreWeave, Crusoe, Latitude.sh.

2
GitHub Actions

Cron every 6h: fetch, deduplicate, commit JSON.

3
Static JSON

Single gpu-data.json (~5MB) on Vercel CDN. No database.

4
Client-Side UI

All filtering/sorting runs in your browser — instant, no server calls.

Coverage

AWS
EC2 GPU instances
Azure
NC, ND, NV series
GCP
A3, A2, N1 GPU
OCI
Bare metal & VM GPU
Lambda Labs
A100, H100, H200
RunPod
Spot & on-demand
Vast.ai
Marketplace spot
Vultr
Cloud GPU
Cudo Compute
On-demand GPU
Nebius
H100, L40S, T4
TensorDock
Marketplace GPUs
DigitalOcean
GPU Droplets
HotAisle
GPU cloud
CloudRift
H100, A100
Verda
GPU cloud
Scaleway
H100, L4 — EU
Hyperstack
H100, A100 — UK/EU
Genesis Cloud
H100, A100 — Iceland
OVHcloud
T4, V100, A100 — EU
CoreWeave
H100, A100 — US
Crusoe
H100 — US (green)
Latitude.sh
Bare metal GPU — global

Disclaimer

Prices may be up to 6 hours old and don't reflect account-specific discounts. Always verify on the provider's site before purchasing.