Hyperstack
Hyperstack provides GPU cloud computing with high-performance NVIDIA GPUs for AI, rendering, and compute workloads.
🇬🇧London, UKFounded2020Developer-firstUpdated 3 hours ago
Cheapest / hr
$0.15
RTX A4000
Most expensive
$7.40
B300
Average / GPU
$1.78
across 47 configs
GPU types
9
47 configurations
VRAM range
16–288GB
per single GPU
Regions
3
with offerings
GPU pricing by type
Median price per GPU/hour across all configurations offered by Hyperstack.
GPU configurations
79 offerings across 9 GPU types. Prices update every 4 hours.
47of 79
| Region | Total VRAM | Storage | Status | ||||||
|---|---|---|---|---|---|---|---|---|---|
16 GB VRAM per GPU· 5 configs | ×1 | 16 GB | 4 | 21 GB | 100 GB | $0.15 | $0.15 | Waitlist | |
48 GB VRAM per GPU· 7 configs | ×1 | 48 GB | 28 | 58 GB | 100 GB | $0.50 | $0.50 | Available | |
48 GB VRAM per GPU· 8 configs | ×1 | 48 GB | 28 | 58 GB | 100 GB + 725 GB | $1.00 | $1.00 | Available | |
NVIDIA A100 80GBTrending 80 GB VRAM per GPU· 6 configs | ×1 | 80 GB | 28 | 120 GB | 100 GB + 750 GB | $1.35–$1.60 | $1.35 | Available | |
96 GB VRAM per GPU· 4 configs | ×1 | 96 GB | 28 | 225 GB | 100 GB + 3.22 TB | $1.85 | $1.85 | Waitlist | |
80 GB VRAM per GPU· 14 configs | ×1 | 80 GB | 28 | 180 GB | 100 GB + 750 GB | $2.50–$3.20 | $2.50 | Available | |
141 GB VRAM per GPU | ×8 | 1.1 TB | 176 | 1.76 TB | 100 GB + 46.88 TB | $3.99 | $31.92 | Available | |
180 GB VRAM per GPU | ×8 | 1.41 TB | 252 | 2 TB | 1,000 GB + 39.06 TB | $6.00 | $48.00 | Available | |
NVIDIA B300Trending 288 GB VRAM per GPU | ×8 | 2.25 TB | 224 | 1.88 TB | 1,000 GB + 39.06 TB | $7.40 | $59.20 | Waitlist |
Similar providers
See incorrect data?
Frequently asked questions
What GPUs does Hyperstack offer?
Hyperstack offers 9 GPU types across 47 configurations, including B300, H100, A100 80GB, B200, H200, L40 and 3 more. The full list with live hourly pricing is shown above.
How much does Hyperstack cost?
Hyperstack GPU pricing starts at $0.15 per GPU-hour, with an average of $1.78 across its on-demand configurations. Spot and reserved options can be cheaper for interruptible or committed workloads.


