Previous-gen data center workhorse for AI training and inference.
There is a 89% difference between the highest and lowest listings for the A100. You can pay up to $5.04/hr, but the market floor is currently $0.52/hr per GPU (on-demand). Spot instances start lower, at $0.20/hr per GPU.
USA 16 configs 1x-8x
USA 1 config 1x
USA 28 configs 1x-8x
USA 9 configs 1x-8x
USA 2 configs 1x
USA 7 configs 1x-8x
USA 4 configs 1x-8x
USA 6 configs 1x-8x
USA 4 configs 1x-8x
USA 20 configs 1x-8x
USA 36 configs 1x-16x
France 15 configs 1x-8x
France 8 configs 1x-8x
Germany 9 configs 1x-8x
Finland 42 configs 1x-8x
Heads up: We do our best to keep these specs & prices accurate. However, cloud costs may fluctuate based on region, usage, and other factors not listed here. These are estimates based on common setups and are for informational purposes only. Always verify current rates & exact specs with the provider before provisioning.
80GB or 40GB HBM2e with 600 GB/s NVLink. The previous-generation workhorse for AI training. Mature software ecosystem and wide cloud availability make it a reliable, well-priced choice.
Lacks FP8 support found in Hopper and Blackwell GPUs. For new large-scale training projects, the H100 offers significantly better performance per dollar. Consider the L40S for pure inference.
The median on-demand price across providers has risen about 5% since August 2025, from $1.70 to $1.79/hr per GPU. This reflects the market-wide median, which moves both when providers change prices and when lower or higher priced offerings enter the market.
With 80GB of VRAM, the A100 is well suited to 30B-class models in FP16, and 70B-class models in 4-bit or 8-bit quantized form.
The A100 has 80GB of VRAM. Multi-GPU setups increase total memory, but that memory is not automatically pooled across GPUs.
The A100 has 1,935 GB/s of memory bandwidth. Higher bandwidth helps with faster data transfer between GPU memory and compute cores.
The A100 supports 7 precision formats. Training: BF16, FP16, TF32, FP32. Inference: INT4, INT8. Scientific: FP64.
Yes. The A100 supports NVLink with 600 GB/s of bidirectional bandwidth. This helps accelerate multi-GPU communication.
A100 pricing currently ranges from $0.20/hr to $5.04/hr per GPU, depending on the provider, instance type, and billing model.
At 720 hours per month, one A100 can cost between $145.08 to $3,628.80 per month, depending on the provider. Reserved and spot pricing can lower that further.
The A100 is available from 41 cloud providers, including Verda, Civo, Runpod. Pricing and availability vary by region and billing model.
Yes. We currently track 371 A100 listings across 41 cloud providers:
| Billing type | Listings | Avg $/GPU/hr |
|---|---|---|
| On-demand | 176 | $1.95/hr |
| Reserved | 146 | $1.75/hr |
| Spot | 48 | $0.76/hr |
| Custom contract | 1 | Custom |
| A100 40GB PCIe | A100 80GB PCIe | A100 40GB SXM | A100 80GB SXM | |
|---|---|---|---|---|
| FP64 | 9.7 TFLOPS | 9.7 TFLOPS | 9.7 TFLOPS | 9.7 TFLOPS |
| FP64 Tensor Core | 19.5 TFLOPS | 19.5 TFLOPS | 19.5 TFLOPS | 19.5 TFLOPS |
| FP32 | 19.5 TFLOPS | 19.5 TFLOPS | 19.5 TFLOPS | 19.5 TFLOPS |
| Tensor Float 32 (TF32) | 156 TFLOPS, 312 TFLOPS* | 156 TFLOPS, 312 TFLOPS* | 156 TFLOPS, 312 TFLOPS* | 156 TFLOPS, 312 TFLOPS* |
| BFLOAT16 Tensor Core | 312 TFLOPS, 624 TFLOPS* | 312 TFLOPS, 624 TFLOPS* | 312 TFLOPS, 624 TFLOPS* | 312 TFLOPS, 624 TFLOPS* |
| FP16 Tensor Core | 312 TFLOPS, 624 TFLOPS* | 312 TFLOPS, 624 TFLOPS* | 312 TFLOPS, 624 TFLOPS* | 312 TFLOPS, 624 TFLOPS* |
| INT8 Tensor Core | 624 TOPS, 1248 TOPS* | 624 TOPS, 1248 TOPS* | 624 TOPS, 1248 TOPS* | 624 TOPS, 1248 TOPS* |
| GPU Memory | 40GB HBM2 | 80GB HBM2e | 40GB HBM2 | 80GB HBM2e |
| GPU Memory Bandwidth | 1,555GB/s | 1,935GB/s | 1,555GB/s | 2,039GB/s |
| Max Thermal Design Power (TDP) | 250W | 300W | 400W | 400W |
| Multi-Instance GPU | Up to 7 MIGs @ 5GB | Up to 7 MIGs @ 10GB | Up to 7 MIGs @ 5GB | Up to 7 MIGs @ 10GB |
| Form Factor | PCIe | PCIe | SXM | SXM |
| Interconnect | NVIDIA® NVLink® Bridge for 2 GPUs: 600GB/s**, PCIe Gen4: 64GB/s | NVIDIA® NVLink® Bridge for 2 GPUs: 600GB/s**, PCIe Gen4: 64GB/s | NVLink: 600GB/s, PCIe Gen4: 64GB/s | NVLink: 600GB/s, PCIe Gen4: 64GB/s |
| Server Options | Partner and NVIDIA-Certified Systems™ with 1-8 GPUs | Partner and NVIDIA-Certified Systems™ with 1-8 GPUs | NVIDIA HGX™ A100-Partner and NVIDIA-Certified Systems with 4, 8, or 16 GPUs, NVIDIA DGX™ A100 with 8 GPUs | NVIDIA HGX™ A100-Partner and NVIDIA-Certified Systems with 4, 8, or 16 GPUs, NVIDIA DGX™ A100 with 8 GPUs |
* With sparsity
** SXM4 GPUs via HGX A100 server boards; PCIe GPUs via NVLink Bridge for up to two GPUs
Source: official Nvidia A100 datasheet.
Ada Lovelace inference GPU. From $0.20/hr per GPU across 32 providers.
Next generation Hopper architecture. From $0.30/hr per GPU across 52 providers.
Consumer Ada Lovelace GPU. From $0.01/hr per GPU across 17 providers.
AMD alternative with 2.4x memory. From $1.45/hr per GPU across 10 providers.
Low-power inference GPU. From $0.12/hr per GPU across 16 providers.
Last updated
