NVIDIA A100 rental: early access
Ampere · SXM (being confirmed)
We’re collecting early-access requests. We’re placing our capacity in Uzbekistan. We don’t publish prices yet. We’ll confirm availability in our reply.
A100
Request A100We’ll reply with terms for your workload. Nothing to pay before the contract.
Estimate usage- Memory per GPU
- 80 GB HBM2e
- Memory bandwidth
- 2.04 TB/s
- GPU interconnect
- NVLink 600 GB/s
NVIDIA A100 specifications
- Architecture
- Ampere
- Announced
- 2020
- Memory
- 80 GB
- Memory type
- HBM2e
- Memory bandwidth
- 2.04 TB/s
- FP32
- 19.5 TFLOPS
- FP16/BF16, dense
- 312 TFLOPS
- FP8, dense
- none
- FP4, dense
- none
- FP64
- 9.7 TFLOPS
- FP64 Tensor Core
- 19.5 TFLOPS
- RT cores
- no
- NVENC/NVDEC
- NVENC 0, NVDEC 5
- MIG
- up to 7 instances
- GPU interconnect
- NVLink 600 GB/s
- Power (TDP/TGP)
- 400 W
- Form factor
- SXM (being confirmed)
Memory is specified per GPU. Multiple GPUs have separate memory: using a larger model across them requires compatible software and model partitioning. NVLink does not make them one GPU.
Other variants: A100 80GB PCIe (opens in a new tab), A100 40GB SXM (opens in a new tab)
The A100 form factor is still being confirmed; memory, its bandwidth and other specifications depend on it. Specs from NVIDIA’s official materials, checked 26 September 2026.
The A100 by workload
- Inference: LLMs up to 32B in BF16 on one GPU
- Fine-tuning: QLoRA up to 70B, LoRA up to 14B on one GPU
- Training from scratch: models up to 3B on one GPU
- Rendering: no RT cores
- FP64 computing: 9.7 TFLOPS, Tensor Core 19.5 TFLOPS
An estimate from open formulas, not a benchmark or an invoice.
Servers with the NVIDIA A100
Server configuration (CPU, RAM, disk, network) is being confirmed.
Sizing assumption: up to 8 GPUs per server. We’ll confirm the actual configuration in our reply.
More than 8 GPUs? Discuss a clusterSources
Similar GPUs
- NVIDIA H100
80 GB HBM3 · 3.35 TB/s · NVLink
H100 leads on 5 of 6 metrics. H100 has more FP32 (67 vs 19.5 TFLOPS) and FP64 Tensor Core throughput (67 vs 19.5 TFLOPS). Equal: memory size.
- NVIDIA RTX PRO 6000 Blackwell
96 GB GDDR7 · 1.6 TB/s · PCIe 5
RTX PRO 6000 leads on 3 of 4 metrics, A100 on 1. RTX PRO 6000 has more FP32 (120 vs 19.5 TFLOPS) and FP16/BF16 (467.8 vs 312 TFLOPS). But A100 has more memory bandwidth (2.04 vs 1.6 TB/s). FP64: 9.7 TFLOPS on A100; not in NVIDIA’s card specs for RTX PRO 6000 (NVIDIA’s whitepaper gives 1/64 of FP32 for the GB202 chip).
The RTX PRO 6000’s FP16, FP8 and FP4 are derived figures. Derived: NVIDIA’s whitepaper per-SM Tensor Core rate × the 2,430 MHz boost clock in the product brief. For the Server Edition, NVIDIA gives FP16/FP8/FP4 only as rounded 1/2/4 PFLOPS peaks, the level of sparse figures; it publishes no dense ones.
We’ll reply with terms for your workload. Nothing to pay before the contract.