Ampere (GA100) architecture • 80 GB HBM2e
NVIDIA A100 (80GB SXM4) hourly pricing
The NVIDIA A100 80GB Ampere GPU remains the foundational workhorse for AI practitioners. Its 80 GB of high-bandwidth HBM2e memory easily accommodates large context windows and multi-gigabyte model weights without memory bottlenecks.
Starting from $1.15/hr on DePIN
Save up to 62% compared to AWS EC2
A100 price per hour across networks
| Network | Network type | Hourly price | Billing type | Status | Audit date |
|---|---|---|---|---|---|
| Akash Network | DePIN (Cosmos) | $1.15 / hr62% less than AWS | dynamic bid | Last checked: 2026-10-01 | 2026-10-01Lease on Akash |
| io.net | DePIN (Solana) | $1.20 / hr61% less than AWS | hourly | Last checked: 2026-09-30 | 2026-09-30Lease on io.net |
| Aethir | DePIN (Arbitrum) | $1.35 / hr56% less than AWS | hourly | Last checked: 2026-09-29 | 2026-09-29Lease on Aethir |
| Amazon Web Services (AWS)Benchmark | Centralized cloud | $3.06 / hr | hourly | Last checked: 2026-10-02 | 2026-10-02Lease on AWS |
| Render Network | DePIN (Solana) | Not verified yetA100 instances on Render are assigned on an ad-hoc basis for enterprise partners; public flat per-hour rate not verified. | job based | Not verified yet | 2026-10-01Lease on Render |
NVIDIA A100 (80GB SXM4) technical specifications
ArchitectureAmpere (GA100)
Total VRAM80 GB HBM2e
Memory bandwidth2.039 TB/s
FP16 compute312 TFLOPS (dense) / 624 TFLOPS (sparse)
FP8 tensor computeN/A (Ampere lacks native FP8 Tensor Cores)
InterconnectNVLink 3.0 (600 GB/s bidirectional)
TDP400W (SXM4) / 250W (PCIe)
Optimal workloads
- •Mid-Scale LLM Fine-Tuning (7B - 70B LoRA/QLoRA)
- •High-Memory Computer Vision & Diffusion Models
- •Enterprise Production Inference
- •Scientific & Molecular Dynamics Simulations
Buyer advice
The A100 provides the best cost-to-stability ratio for teams running LoRA fine-tuning and production serving of models up to 34B parameters. At $1.15–$1.35/hr on decentralized clouds versus $3.06/hr on AWS EC2 p4d.24xlarge, decentralized A100s offer savings of over 55%.
Disclosure: Some links on this page may be affiliate referral links. When you lease compute through these links, DecentralGPU may earn a small referral commission at no additional cost to you. See our affiliate disclosure and pricing methodology.