NVIDIA · Ampere · 48 GB GDDR6 · PCIe · 2020
3 providers · 4 configurations · prices as of Sep 27, 2026, 01:12 UTC
Daily A40 prices per GPU-hour. The shaded period is reconstructed from the version history of SkyPilot's public price catalogs (fewer providers than live data); after that, PetaPrice's own tracking.
Cheapest On-demand A40 offer at each provider, per GPU-hour. Shorter and greener is cheaper.
Thin vertical line = median across providers ($0.49).
3 providers · 4 configurations
GPUs with similar compute and memory, and their lowest on-demand price right now.
On-demand A40 prices start at $0.49 per GPU-hour (RunPod). The median across 2 providers is $0.49, and the most expensive is $0.49. Running one GPU around the clock costs about $358 per month at the lowest price.
Yes. Spot (interruptible) A40 capacity starts at $0.40 per GPU-hour, 18% below the cheapest on-demand price. Spot instances can be preempted at any time, so use them for checkpointed training or batch jobs.
At $0.49 per hour and 0.15 PF of dense FP16/BF16, the A40 costs $3.27 per PFLOP-hour. That ranks 49 of 70 GPUs with on-demand prices on PetaPrice. Figures are datasheet peaks; real utilization is typically 30–60%.
48 GB GDDR6 with 696 GB/s of memory bandwidth, Ampere architecture (2020), PCIe form factor, 300 W. Dense peak: 0.15 PF FP16 / BF16, 0.299 PF INT8, 0.0748 PF TF32, 0.0374 PF FP32, 0.0006 PF FP64.
Prices were collected on Sep 27, 2026, 01:12 UTC from provider APIs and public catalogs and refresh about every 2 hours. They are indicative list prices; always confirm on the provider's website before booking.