V100 GPU rental price
NVIDIA V100 · 16 GB VRAM. Live per-GPU pricing aggregated across 3 providers and 3 regions.
The V100 (16GB or 32GB HBM2) currently rents from $0.080 per GPU per hour on Aquanode, across 3 providers.
The cheapest V100 offer right now (SimplePod, $0.080/GPU/hr) is about 58% below the market median of $0.190/GPU/hr.
V100 specs
V100 full specs
| Architecture | NVIDIA, launched 2017 |
| VRAM | 16GB or 32GB HBM2 |
| Memory bandwidth | 900 GB/s |
| FP16 / BF16 tensor throughput | 125 TFLOPS (SXM2), 112 TFLOPS (PCIe) (peak, dense) |
| Interconnect | NVLink, 300 GB/s (SXM2 only) |
| TDP | 300W (SXM2), 250W (PCIe) |
| Form factor | SXM2, PCIe, full height/length |
Specs sourced from the vendor's public datasheet. See the source.
What fits in 16GB or 32GB HBM2 of VRAM
| Model | Precision | Fits? |
|---|---|---|
| Llama 3 8B | FP16 | roughly 16GB. Fits on the 32GB card, not the 16GB one. |
| Mistral 7B | FP16 | ~14GB. Fits on either card, tightly on the 16GB one. |
| Llama 3.1 70B | FP16 | ~140GB. Needs 5+ cards even at 32GB each. |
| Llama 3.1 70B | INT4 | ~35-40GB. The INT4 kernels do not run on Volta at all, so this is not an option regardless of VRAM. |
Approximate, based on published parameter counts and standard bytes-per-parameter rules of thumb (FP16 ≈ 2 bytes/param, INT4 ≈ 0.5-0.6 bytes/param). Real footprint also depends on KV-cache size and framework overhead.
Good for
HBM2 bandwidth (900 GB/s) at the cheapest hourly rate of any HBM card on this marketplace, and the SXM2 version has real NVLink at 300 GB/s. For FP16 training and classical HPC it is still a lot of memory bandwidth per dollar.
Not good for
No BF16, no FP8, and compute capability 7.0. Below the 7.5 floor AWQ/GPTQ/Marlin INT4 kernels require, so quantized serving does not run on it at all. Most current LLM training recipes assume BF16, which means a V100 needs an FP16 recipe with loss-scaling or it does not run.
Get notified when the price drops
GPU supply moves hourly. Tell us what you're waiting for and we'll email you when a matching offer appears across any provider we track.
How this price is calculated
All prices on this page are normalized to a per-GPU hourly rate using each offer's authoritative GPU count, so that raw price is divided by the number of GPUs it actually covers; Akash reports its price as already per-GPU, so it is used as-listed. An offer with a missing, zero, or invalid GPU count is excluded entirely rather than published at a guessed rate.
No offers were excluded from this snapshot for a missing or invalid price. No offers were dropped as price outliers in this snapshot.
Only the cheapest qualifying offer per provider is shown in the table above. This page regenerates at most once per hour.
Frequently asked questions
How much does it cost to rent a V100?
Live V100 rental prices currently range from $0.080 to $0.230 per GPU per hour across 3 providers, with a median of $0.190 per GPU per hour.
Which provider has the cheapest V100?
SimplePod currently offers the lowest V100 rate on Aquanode's marketplace at $0.080 per GPU per hour in US.
How is the V100 price calculated?
All prices are normalized to a per-GPU hourly rate using each offer's authoritative GPU count, which the raw price is divided by; Akash reports its price as already per-GPU. Offers whose price can't be safely normalized, or whose rate is an extreme outlier against the rest of the market, are excluded.
How much does an V100 cost per hour?
Live V100 rental rates on Aquanode currently range from $0.080 to $0.230 per GPU per hour, with a median of $0.190/hr. See the live table above for current per-provider pricing.
How much VRAM does an V100 have?
The V100 has 16GB or 32GB HBM2, with 900 GB/s of peak memory bandwidth.
Is renting cheaper than buying?
Renting avoids the upfront hardware cost and lets you match spend to actual usage. A rented V100 at $0.080/hr only costs money while it's running, whereas buying ties up capital in hardware that keeps depreciating whether it's in use or not. Which is cheaper depends on how continuously you'd run it; short or bursty workloads usually favor renting.
