The NVIDIA V100 was once the gold standard for AI workloads. But in 2026, it's in an awkward spot: widely available at very low prices, but outperformed by newer GPUs.
Today, cloud providers rent V100 GPUs for as little as $0.14/hour. At first glance, this looks like a great deal. However, modern alternatives like the RTX A6000 are available for $0.35/hour on Thunder Compute, offering a massive leap in performance.
Takeaways
<ul><li>The <strong>NVIDIA V100 price</strong> ranges from <strong>$0.14/hr to over $3/hr</strong> depending on the provider.</li><li>Marketplace platforms offer significantly lower pricing but with tradeoffs in availability and reliability.</li><li>Hyperscalers charge a premium for legacy infrastructure.</li><li>Despite low entry pricing, <strong>performance per dollar is no longer competitive</strong>.</li><li>Modern GPUs deliver substantially better throughput at similar or slightly higher costs.</li></ul>
| Provider | V100 Price (Hourly) | VRAM | Instance Name |
|---|---|---|---|
| Verda | $0.14 | 16GB | N/A |
| TensorDock | $0.19 | 16GB | N/A |
| Vast.ai | $0.25 | 16GB | N/A |
| Lambda | $0.79 | 16GB | N/A |
| Paperspace | $2.30 | 32GB | N/A |
| Google Cloud | $2.50 | 16GB | n1-standard |
| AWS | $3.06 | 16GB | p3.2xlarge |
| Azure | $3.34 | 16GB | Standard_NC6s_v3 |
| * Prices reflect on-demand hourly rates as of May 2026. Availability of specific VRAM variants may vary by region. | |||
Methodology (why you can trust these numbers)
<ul><li>Pricing reflects <strong>public on-demand rates</strong> as of May 2026.</li><li>Only <strong>comparable instance types</strong> (single GPU, standard configurations) were included.</li><li>Marketplace pricing reflects <strong>live listings</strong>, which may fluctuate.</li><li>Hyperscaler pricing reflects <strong>standard on-demand rates</strong>, excluding discounts.</li></ul>
Why this matters for developers
This chart compares how much it costs to run an NVIDIA V100 for 10 hours across major cloud providers.
Thunder Compute offers NVIDIA RTX A6000 GPUs for $0.35/hr which have 3 times the VRAM as NVIDIA V100.
Choosing outdated GPUs like the V100 can result in higher cost per unit of compute, even if hourly rates appear lower.
AWS P3 Instances
AWS offers NVIDIA V100 GPUs in its EC2 P3 Instances with:
<ul> <li> NVIDIA V100 GPUs <ul> <li> 16GB VRAM on standard P3 <li> 32GB VRAM on P3dn </ul> <li> 8-96 vCPUs <li> 61-768GB RAM <li> 10-100 Gbps networking throughput</ul>
As for all AWS instances, egress fees are $0.09/GB after the first free 100GB.
| Provider | SKU | V100 GPUs | VRAM | VCPUs | RAM | Hourly price | Price per-GPU |
|---|---|---|---|---|---|---|---|
| AWS | p3.2xlarge | 1 | 16 | 8 | 61 | $3.06 | |
| AWS | p3.8xlarge | 4 | 16 | 32 | 244 | $12.24 | $3.06 |
| AWS | p3.16xlarge | 8 | 16 | 64 | 488 | $24.48 | $3.06 |
| AWS | p3dn.24xlarge | 8 | 32 | 96 | 768 | $31.21 | $3.90 |
Azure ND V100 Instances
Through the NDv2-series, Azure offers high-performance virtual machines designed for demanding deep learning training and HPC workloads.
Powered by NVIDIA Tesla V100 NVLink GPUs and Intel Xeon Platinum 8168 processors, these instances feature interconnected GPUs via NVLink for massive multi-GPU scaling performance.
| Provider | SKU | V100 GPUs | VRAM | VCPUs | RAM | Hourly price | Price per-GPU |
|---|---|---|---|---|---|---|---|
| Azure | Standard_ND40rs_v2 | 8 | 32 | 40 | 672 | $22.03 | $2.75 |
Google Cloud N1 instances (V100)
Through general-purpose N1 machine types, GCP allows custom attachments of NVIDIA V100 GPUs in standard, highmem, and highcpu configurations.
Being a highly customizable instance, the table below is just illustrative of the smallest and largest machine types from each configuration that support V100 configurations.
| Provider | SKU | V100 GPUs | VCPUs | RAM | SSD Storage | Hourly price | Price per-GPU |
|---|---|---|---|---|---|---|---|
| GCP | n1-standard-8 | 1 | 8 | 30 | 1 x 375GB | $2.90 | |
| GCP | n1-standard-96 | 8 | 96 | 360 | 1 x 375GB | $25.39 | $3.17 |
| GCP | n1-highcpu-8 + 1xV100 | 1 | 8 | 7.2 | 1 x 375GB | $2.81 | |
| GCP | n1-highcpu-96 + 8xV100 | 8 | 96 | 86.4 | 24 x 375GB | $24.23 | $3.03 |
| GCP | n1-highmem-8 + 1xV100 | 1 | 8 | 52 | 1 x 375GB | $3.00 | |
| GCP | n1-highmem-96 + 8xV100 | 8 | 96 | 624 | 1 x 375GB | $26.51 | $3.31 |
Last Thoughts on NVIDIA V100 Pricing
The NVIDIA V100 still offers low entry pricing, but it is no longer a strong default choice.
For most workloads, newer GPUs deliver significantly better performance per dollar, making them the better long-term investment.
Unless you're extremely cost-constrained, it's worth upgrading to a modern GPU like the RTX A6000.
FAQ
Is the NVIDIA V100 still worth it in 2026?
Only for very budget-constrained workloads. For most use cases, newer GPUs provide significantly better performance per dollar.
Why is the V100 so cheap now?
It’s widely available and has been replaced by newer architectures like Ampere and Hopper, driving prices down.
What is the best alternative to the V100?
The RTX A6000 is one of the best alternatives, offering a large performance boost for a relatively small increase in cost.
How much faster is A6000 vs V100?
In many AI workloads, the A6000 can deliver 2–4x better performance depending on optimization and precision.