What AWS, Azure, Google Cloud, RunPod, Lambda and CloudRift charge per GPU-hour, read from each provider's own price sheet and divided down to the single card. On demand, Linux, US regions, no commitment.
Provider prices read October 15, 2026. CloudRift column live, read October 16, 2026 at 6:10 PM UTC.
One number per cell, with the instance shape it comes from underneath. Multi-GPU shapes are divided by their GPU count. Where a provider has no on-demand shape for a card, the cell says so instead of guessing.
| GPU | AWS | Azure | Google Cloud | RunPod | Lambda | CloudRift |
|---|---|---|---|---|---|---|
| NVIDIA H10080 GB | $6.88p5.48xlarge · 8 GPUs | $12.29ND96isr_H100_v5 · 8 GPUsNVL 94 GB single GPU: $6.98 | $11.06a3-highgpu-8g · 8 GPUs | $3.99H100 SXMPCIe: $2.89 | $3.998x H100 SXM · 8 GPUssingle GPU: $4.29 | not offered |
| NVIDIA H200141 GB | $7.91p5en.48xlarge · 8 GPUs | $10.60ND96isr_H200_v5 · 8 GPUsEast US 2 | $10.60a3-ultragpu-8g · 8 GPUs | $5.29H200 SXM | not offered | not offered |
| NVIDIA B200180 GB | $14.24p6-b200.48xlarge · 8 GPUs | $27.04ND128isr_NDR_GB200_v6 · 4 GPUsGB200 NVL72 slice | $8.06a4-highgpu-8g · 8 GPUsFlex-start only, no on-demand price | $7.99B200 | $6.698x B200 SXM6 · 8 GPUssingle GPU: $6.99 | not offered |
| NVIDIA A10080 GB | $3.43p4de.24xlarge · 8 GPUs | $4.10ND96amsr_A100_v4 · 8 GPUsPCIe single GPU: $3.67 | $5.07a2-ultragpu-1g | $1.79A100 SXM | $2.798x A100 SXM · 8 GPUs | $1.0580 GB, fully booked right now |
| NVIDIA A10040 GB | $2.74p4d.24xlarge · 8 GPUs | $3.40ND96asr_v4 · 8 GPUs | $3.67a2-highgpu-1g | not offered | $1.991x A100 PCIe | not offered |
| NVIDIA L40S48 GB | $1.86g6e.xlarge8 GPUs: $3.77 each | not offered | not offered | $1.09L40S | not offered | $0.6348 GB, available now |
| NVIDIA V10016 GB | $3.06p3.2xlarge | $3.06NC6s_v3 | $2.48N1 + V100GPU only, VM billed on top | not offered | $0.798x Tesla V100 · 8 GPUs | $0.2516 GB $0.25, 32 GB $0.28 |
| NVIDIA T416 GB | $0.53g4dn.xlarge | $0.53NC4as_T4_v3 | $0.35N1 + T4GPU only, VM billed on top | not offered | not offered | not offered |
| AMD Instinct MI300X192 GB | not offered | $6.00ND96is_MI300X_v5 · 8 GPUsEast US 2 | not offered | not offered | not offered | $4.00MI350X 288 GB, newer generation, fully booked right now |
Rates, instance sizes and free capacity below come from the CloudRift API each time this page is rebuilt, so a card that is fully booked says so rather than quoting a price you cannot start today.
| GPU | On Demand | Cheapest Elsewhere | Per Instance | Right Now |
|---|---|---|---|---|
| NVIDIA V100 SXM216 GB | $0.25per GPU-hour | Lambda $0.79 in 8-GPU boxes, AWS and Azure $3.06 | up to 8 GPUs | Fully bookedon demand when a node frees up |
| NVIDIA V100 SXM332 GB | $0.28per GPU-hour | Lambda $0.79 in 8-GPU boxes, AWS and Azure $3.06 | up to 16 GPUs | Available now1 to 10 GPUs per instance · United States |
| NVIDIA RTX 409024 GB | $0.39per GPU-hour | RunPod $0.89, community tier $0.34 | up to 10 GPUs | Fully bookedon demand when a node frees up |
| NVIDIA RTX 509032 GB | $0.62per GPU-hour, multi-GPU boxes from $0.60 | RunPod $1.19, community tier $0.69 | up to 8 GPUs | Available nowsingle-GPU instances · South Korea |
| NVIDIA L40S48 GB | $0.63per GPU-hour | RunPod $1.09, AWS $1.86 | up to 8 GPUs | Available now1 to 7 GPUs per instance · Taiwan |
| NVIDIA A100 SXM480 GB | $1.05per GPU-hour | RunPod $1.79, AWS $3.43 in 8-GPU boxes | up to 8 GPUs | Fully bookedon demand when a node frees up |
| NVIDIA RTX PRO 600096 GB | On requestreserved, from $1.16 per GPU-hour | RunPod $2.49 | up to 8 GPUs | Fully bookedquoted on request |
| NVIDIA RTX PRO 6000 Max-Q96 GB | On requestreserved, from $1.40 per GPU-hour | RunPod $2.49 | up to 8 GPUs | Fully bookedquoted on request |
| AMD MI350X288 GB | $4.00per GPU-hour | MI300X on Azure and Oracle $6.00 | up to 8 GPUs | Fully bookedon demand when a node frees up |
An hourly rate says little until you know what the card does in that hour. These are the published benchmarks behind the comparisons above, with the pricing basis each one used. We have no V100 or MI350X benchmark yet, so those rows stand on price alone.
28% lower cost per token on single-GPU inference. Once a model needs 8-way tensor parallelism over NVLink, the H100 and H200 pull 3 to 4 times ahead. That test priced the cards at RunPod rates, not ours.
Read the benchmarkOn a 24 GB-class model and one GPU: $0.048, $0.040 and $0.043 per million tokens. The 5090 is the cheapest per token at that size, and the PRO 6000 takes over as models grow.
Read the benchmarkUp to 4.9 times the throughput of the RTX PRO 6000 and the best cost efficiency in the set. That comparison used estimated run costs rather than cloud list prices, which is why it is not in the table above.
Read the benchmarkWhy two honest comparisons of the same GPU can disagree by a factor of two.
Pricing FAQ
The console shows the same hourly rates as this page, with live availability per instance size. Reserved terms for the cards above are quoted on request.