NVIDIA A100 SXM: Specs, Pricing & Cloud Availability
Last updated: 2026-03-22
Technical Specifications
| Architecture | Ampere |
|---|---|
| VRAM | 80 GB HBM2e |
| Memory Bandwidth | 2.0 TB/s |
| FP16 Performance | 624 TFLOPS |
| FP8 Performance | 624 TFLOPS |
| TDP | 400W |
| Interconnect | NVLink 3.0 |
Cloud Pricing
| Provider | On-Demand $/hr | Spot $/hr | Availability |
|---|---|---|---|
| Microsoft Azure | $3.67/hr | N/A | Available |
| RunPod | $1.99/hr | N/A | Available |
| Lambda Labs | $1.29/hr | N/A | Available |
| CoreWeave | $2.21/hr | N/A | Available |
| Together AI | $2.00/hr | N/A | Available |
| Vast.ai | $1.60/hr | N/A | Available |
| Vultr | $2.85/hr | N/A | Available |
| Oracle Cloud (OCI) | $3.50/hr | N/A | Available |
| Cudo Compute | $1.80/hr | N/A | Available |
| FluidStack | $1.75/hr | N/A | Available |
| Paperspace (DigitalOcean) | $3.18/hr | N/A | Available |
Benchmarks
The NVIDIA A100 SXM delivers 624 TFLOPS FP16 and 312 TFLOPS FP32 performance with 2.0 TB/s memory bandwidth.
Best Use Cases
The NVIDIA A100 SXM is optimized for Production ML training and inference.