For investors in AI compute
Real-time GPU
availability index.
Track H100, H200, and B200 GPU availability across major cloud providers.
Current Availability
-
Latest poll result
Daily Availability
-
Today's average
30-Day Average
-
Trailing 30-day average
Availability over time
% of checks where at least one provider had capacity available
H100 30-day availability by provider
Individual provider trends
Provider status
Stale if no update in 15 minutes
| Provider | Status | H100 | H200 | B200 | Last updated |
|---|---|---|---|---|---|
| Loading... | |||||
What it measures
Percentage of checks where at least one GPU instance was available to rent across tracked providers. Aggregated into daily availability scores.
How to use it
High availability means capacity is easy to find. Low availability signals tight supply. The 30-day moving average smooths out spikes to show the underlying trend.
Methodology
Providers tracked
- RunPod
- Vast.ai
- Lambda
GPU types tracked
- H100 80GB
- H200 141GB
- B200
How availability is defined
A GPU counts as available if at least one rentable single-GPU instance existed at the time of the check, priced under $4/hr for H100 and H200. B200 has no price filter.
Data collection
Polled every 5 minutes. Availability is calculated per provider then averaged across providers. Daily availability and 30-day moving average are calculated from raw checks.
Limitations
- Region data not available for all providers
- Availability reflects API responses, not physical hardware
- H200 historical data not available prior to deployment date