Quick Insights
Performance/Dollar
4.19 TFLOPS/$k
FP16 performance per $1000
VRAM/Dollar
2.5 GB/$k
VRAM per $1000
vs Data Center Average
-25% perf
FP16 TFLOPS comparison
Cloud Availability
14 providers
from $1.47/hr
Specifications
| VRAM | 80GB HBM3 |
| Memory Bandwidth | 3.4 TB/s |
| FP16 TFLOPS | 134.0 |
| Tensor TFLOPS | 2.0k |
| FP32 TFLOPS | 67.0 |
| TDP | 700W |
| Form Factor | SXM |
| Architecture | Hopper |
| NVLink | Yes (900GB/s) |
| Release Date | 2022-09 |
Buy vs Rent Analysis
Buy Hardware
$32k
- One-time cost, unlimited usage
- Full control over hardware
- Electricity & cooling costs extra
- Depreciation over 2-3 years
Best if using >21801 hours total
Rent Cloud GPU
$1.47/hr
- Pay only for what you use
- No upfront investment
- Scale up/down instantly
- No maintenance required
Best for <21801 hours or variable usage
Breakeven Point
21,801
hours of usage
At $1.47/hr cloud pricing, buying the hardware pays off after 21,801 hours (~908 days or 30.3 months of 24/7 usage).
| Usage | Monthly Cloud Cost | Months to Breakeven |
|---|---|---|
| 100 hrs/month | $146.78 | 219 months |
| 200 hrs/month | $293.56 | 110 months |
| 500 hrs/month | $733.90 | 44 months |
Cloud GPU Pricing
Rent H100 from 14 cloud providers. Prices shown per GPU per hour.
| Provider | Type | Instance | GPUs | On-Demand | Per GPU | Spot | Availability |
|---|---|---|---|---|---|---|---|
| Vast.ai | marketplace | vastai-h100-sxm | 1x | $1.47/hr | $1.47/hr Cheapest | $1.47/hr | - |
| Fluidstack | gpu-cloud | fluidstack-h100-sxm | 1x | $2.10/hr | $2.10/hr | - | - |
| Genesis Cloud | gpu-cloud | genesis-h100-sxm | 1x | $2.19/hr | $2.19/hr | - | - |
| Paperspace | gpu-cloud | paperspace-h100-sxm | 1x | $2.24/hr | $2.24/hr | - | - |
| TensorDock | marketplace | tensordock-h100-sxm | 1x | $2.25/hr | $2.25/hr | - | - |
| Datacrunch (Verda) | gpu-cloud | datacrunch-h100-sxm | 1x | $2.29/hr | $2.29/hr | - | - |
| Jarvis Labs | gpu-cloud | jarvislabs-h100-sxm | 1x | $2.99/hr | $2.99/hr | - | - |
| RunPod | gpu-cloud | NVIDIA H100 80GB HBM3 | 1x | $3.29/hr | $3.29/hr | $3.29/hr | - |
| Lambda Labs | gpu-cloud | lambda-h100-sxm | 1x | $3.99/hr | $3.99/hr | - | - |
| CoreWeave | gpu-cloud | coreweave-h100-sxm | 1x | $6.16/hr | $6.16/hr | - | - |
| Google Cloud Platform | hyperscaler | gcp-h100 | 1x | $9.80/hr | $9.80/hr | $2.25/hr (-77%) | - |
| Google Cloud Platform | hyperscaler | gcp-h100_mega | 1x | $10.34/hr | $10.34/hr | $2.38/hr (-77%) | - |
| Oracle Cloud | hyperscaler | oracle-gpu-h100 | 1x | $10.75/hr | $10.75/hr | - | - |
| Microsoft Azure | hyperscaler | Standard_ND96isr_H100_v5 | 8x | $98.32/hr | $12.29/hr | - | - |
Best Spot Deal:
Google Cloud Platform
offers spot pricing at $2.25/hr
(77% off on-demand).
H100 vs Alternatives
Compare H100 with similar GPUs from other brands.
| GPU | VRAM | FP16 TFLOPS | Bandwidth | Hardware Price | Cloud Price | |
|---|---|---|---|---|---|---|
| H100 Current | 80GB | 134.0 | 3.4 TB/s | $32k | - | - |
| AMD Instinct MI210 AMD | 64GB (-20%) | 181.0 (+35%) | 1.6 TB/s | - | - | Compare |
| MI300 AMD | 128GB (+60%) | 490.3 (+266%) | 5.3 TB/s | $15k (-53%) | - | Compare |
| AMD Instinct MI300A AMD | 128GB (+60%) | 980.0 (+631%) | 5.3 TB/s | - | - | Compare |
| AMD Instinct MI250X AMD | 128GB (+60%) | 383.0 (+186%) | 3.3 TB/s | - | - | Compare |
| AMD Instinct MI250 AMD | 128GB (+60%) | 362.0 (+170%) | 3.3 TB/s | - | - | Compare |
Compare H100
Alternatives
- AMD Instinct MI210 64GB · -
- MI300 128GB · $15k
- AMD Instinct MI300A 128GB · -
- AMD Instinct MI250X 128GB · -
- AMD Instinct MI250 128GB · -
