ultraData Center
A100 SXM 40GB GPU
The 40 GB version of the Ampere A100 in the SXM4 form factor, with HBM2 memory at 1,555 GB/s and NVLink for multi-GPU nodes. It is the A100 in AWS p4d, Azure ND96asr_v4 and Google a2-highgpu instances.
VRAM 40GB
CUDA Cores 6,912
Tensor Cores 432
TDP 400W
Process 7nm
From · On-demand
$0.840/hr
across 14 providers
Cloud Pricing
Cheapest on-demand on Akash Network — 58% below avg| Provider | Price / hr |
|---|---|
$0.840/hr 1× | |
$0.843/hr 2× | |
$0.845/hr 1× | |
$0.846/hrSpot 1×8× | |
$1.00/hr 1× | |
$1.15/hr 8× | |
$1.22/hrReserved · 36mo 16× | |
$1.29/hrReserved · 36mo 1×2×4×8× | |
$1.40/hr 1×8× | |
$1.47/hr 1×8× | |
$1.50/hr 1× | |
$1.50/hr 1× | |
$1.54/hr 8× | |
$1.54/hr 1×8× | |
$1.69/hr 1× | |
$1.69/hr 8× | |
$1.99/hr 1× | |
$1.99/hr 1×8× | |
$1.99/hr 1× | |
$2.08/hrSpot 16× | |
$2.19/hrSpot 1×2×4×8× | |
$2.19/hr 1× | |
$2.19/hrReserved · 12mo 16× | |
$2.31/hrReserved · 12mo 1×2×4×8× | |
$2.74/hr 8× | |
$3.40/hr 8× | |
$3.48/hr 16× | |
$3.67/hr 1×2×4×8× |
Prices updated daily. Last check: Nov 14, 2026
A100 SXM 40GB pricing by provider
Every configuration, price rank, and alternative for one provider at a time.
Performance
FP16
312 TFLOPS
FP32
19.5 TFLOPS
BF16
312 TFLOPS
INT8
624 TOPS
Bandwidth
1555 GB/s
Common Use Cases
Multi-GPU training and fine-tuning of models that fit in 40 GB per GPU, batch inference, HPC workloads using FP64, and MIG partitioning into up to seven instances
Full Specifications
Hardware
- Manufacturer
- NVIDIA
- Architecture
- Ampere
- CUDA Cores
- 6,912
- Tensor Cores
- 432
- RT Cores
- 0
- Process Node
- 7nm
- TDP
- 400W
Memory & Performance
- VRAM
- 40GB
- Memory Interface
- 5120-bit
- Memory Bandwidth
- 1555 GB/s
- FP32
- 19.5 TFLOPS
- FP16
- 312 TFLOPS
- BF16
- 312 TFLOPS
- FP64
- 9.7 TFLOPS
- INT8
- 624 TOPS
- Release
- 2020