Previous-gen data center workhorse for AI training and inference.
Weekly median price per GPU per hour · Get the data
By provider shows one card per company, placed where its best offer ranked. By configuration lists every offer.
What you can rent comes first: in stock, then waitlist, then not reported, then out of stock. Priced offers rank ahead of quote-only, and on-demand ahead of other billing types.
Within a group, five factors set the order:
We match the provider, its country, GPU model, form factor, billing type, availability and instance name. Matching is partial and case-insensitive. Everyday words work too, so "interruptible" finds spot and "sold out" finds out of stock.
No providers match .
Heads up: A provider's own page may quote a different figure, for example on tax, region or a promotion. A monthly-only plan shows a derived hourly rate. Verify before provisioning. More on how we price.
One A100 has 80 GB of VRAM on the PCIe variant. In practice, that's enough memory for roughly 113B parameters at 4-bit or 31B at 16-bit, assuming a 32K context. Below are some open-weight LLMs, with the estimated memory and GPUs each one needs.
| Model | Memory (INT4 / FP4) | A100s needed | Cost /hr | Cost /mo |
|---|---|---|---|---|
Qwen3.8-27B
27B
|
19 GB
|
1
|
$1.88
|
$1,354
|
Gemma 4 31B
30.7B
|
24 GB
|
1
|
$1.88
|
$1,354
|
GPT-OSS-120B
117B · 5.1B active
|
68 GB
|
1
|
$1.88
|
$1,354
|
GLM-5.3-Flash
320B · 18B active
|
178 GB
|
4
|
$7.52
|
$5,414
|
MiniMax-M3
428B · 23B active
|
241 GB
|
4
|
$7.52
|
$5,414
|
DeepSeek V4.1 Flash
552B · 16B active
|
306 GB
|
8
|
$15.04
|
$10,829
|
Kimi K3
2.8T · 104B active
|
1,545 GB
|
22
|
–
|
–
|
Estimates based on the median on-demand rate. Memory is weights plus FP16/BF16 KV cache at 32K context per request. GPU counts assume 90% of advertised VRAM is usable. No guarantee of runtime support, usable performance, or that a matching quantized build exists. Pricing methodology.
Nvidia A100 · Per GPU · PCIe
| Compute · dense | |
|---|---|
| FP16 / BF16 | 312 TFLOPS 624 with sparsity |
| INT8 | 624 TOPS 1,248 with sparsity |
| FP32 | 19.5 TFLOPS |
| FP64 | 9.7 TFLOPS |
| Precision support | FP16BF16TF32FP32FP64INT4INT8 |
| Memory | |
|---|---|
| Capacity | 80 GB HBM2e |
| Bandwidth | 1,935 GB/s |
| Bus width | 5,120-bit |
| ECC | Yes |
| Silicon | |
|---|---|
| Architecture | Ampere |
| Process | TSMC N7 |
| Transistors | 54.2 billion |
| Shader cores | 6,912 CUDA cores |
| Matrix cores | 432 Tensor cores |
| Compute units | 108 SMs |
| Fabric and host | |
|---|---|
| GPU interconnect | NVLink 600 GB/s |
| Host interface | PCIe 4.0 x16 |
| Power | |
|---|---|
| Board power | 300 W |
| Cooling | Passive |
| Platform | |
|---|---|
| Partitioning | MIG, up to 7 instances |
Source: official Nvidia A100 datasheet.
The A100 ships in 4 versions with different memory, bandwidth or interconnect. The figures at the top of this page are for the A100 PCIe.
| Variant | Memory | Memory bandwidth | Interconnect |
|---|---|---|---|
A100 PCIe | 80 GB | 1,935 GB/s | NVLink 600 GB/s |
A100 SXM | 80 GB | 2,039 GB/s | NVLink 600 GB/s |
A100 SXM | 40 GB | 1,555 GB/s | NVLink 600 GB/s |
A100 PCIe | 40 GB | 1,555 GB/s | NVLink 600 GB/s |
As of September 23, 2026, the median on-demand price is $1.88 per GPU per hour across 37 providers with a priced on-demand config.
| Billing type | Configs | Median /GPU/hr | Cheapest |
|---|---|---|---|
On-demand | 215 | $1.88 | $0.43 (in stock, Vast.ai) |
Reserved | 202 | $1.43 | $0.66 (24 mo) |
Spot | 46 | $1.14 | $0.19 (in stock, Vast.ai) |
2 providers also quote the A100 on a custom contract, priced per deal.
At 720 hours per month, one A100 costs an estimated $1,354 at the median on-demand price. Cheapest verified in stock: $310 per month on-demand, $533 reserved, $137 spot.
GetDeploying currently tracks A100 configs from 47 providers. The cheapest verified in-stock on-demand configs come from Vast.ai, GPU.ai, Jarvislabs and Lium. See the full price comparison above for every provider and config.
As of September 23, 2026, the median on-demand price has fallen about 4% over the past 90 days to $1.88 per GPU per hour, though it is about 7% above where it was a year ago.
One A100 runs models up to roughly 113B parameters at 4-bit quantization or 31B at 16-bit, assuming a 32K context. Larger models run across multiple GPUs: the model table above shows the estimated memory and GPU count for popular open-weight LLMs.
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.
Nvidia H100
80 GB HBM3 · 73% more bandwidth than the A100
Nvidia RTX 5090
32 GB GDDR7 · PCIe only
Nvidia RTX 4090
24 GB GDDR6X · 48% less bandwidth than the A100
AMD MI300X
192 GB HBM3 · 2.7x the bandwidth of the A100
Nvidia L40S
48 GB GDDR6 · 55% less bandwidth than the A100
Last updated
