NVIDIA · Ampere Architecture

Rent NVIDIA A100 SXM (40GB) in the Cloud

Budget A100 variant with 40GB VRAM. Good for training models under 13B parameters and batch inference.

VRAM 40 GB HBM2e
Bandwidth 1,555 GB/s
FP16 312.0 TFLOPS
FP32 19.5 TFLOPS
TDP 400W
Architecture Ampere
Cheapest On-Demand $0.80/hr
Average On-Demand $0.97/hr
Providers 2

Compare NVIDIA A100 SXM (40GB) Cloud Pricing — 2 Providers

On-Demand

NVIDIA A100 SXM (40GB) — On-Demand pricing
Provider Price / GPU / hr Availability Notes Action
Vast.ai United States $0.80/hr CHEAPEST Available Marketplace avg Visit Provider
RunPod United States $1.14/hr Available Secure Cloud, SXM Visit Provider

Prices last verified: April 13, 2026

NVIDIA A100 SXM (40GB) Technical Specifications

Manufacturer NVIDIA
Architecture Ampere
VRAM 40 GB HBM2e
Memory Bandwidth 1,555 GB/s
FP16 (Tensor) 312.0 TFLOPS
FP32 19.5 TFLOPS
TDP 400W
Release Year 2020
Segment Data center
Memory Type HBM2e

Best For

AI training fine-tuning inference on smaller models

Compare with Other GPUs

See how NVIDIA A100 SXM (40GB) stacks up against other popular cloud GPUs in specs, pricing, and availability.