GPU Pricing Index

Rank cloud GPUs by normalized rental cost, provider coverage, spot discount, and budget tier. Use the index to choose a starting GPU before opening a detailed provider page or rental calculator.

60priced GPU families
14tracked providers
$0.060index baseline / GPU-hour

Last pricing refresh: Jan 15, 2026. Index score 100 equals the lowest tracked on-demand GPU-hour in the current dataset.

Compare GPU price index scores

Filter by budget tier, sort by index score, and jump to the matching GPU or calculator.

Index GPU Lowest on-demand Median on-demand Spot discount Providers Tier Next step
100 RTX A4000
16GB VRAM
$0.060/hr $0.320/hr -167% 4 Value Estimate · Details
134 RTX 3090
24GB VRAM
$0.081/hr $0.200/hr 23% 3 Balanced Estimate · Details
217 RTX 3070
8GB VRAM
$0.130/hr $0.130/hr 46% 1 Balanced Estimate · Details
233 V100
32GB VRAM
$0.140/hr $0.800/hr -609% 5 Balanced Estimate · Details
283 NVIDIA GeForce RTX 3080
10GB VRAM
$0.170/hr $0.170/hr 47% 1 Premium Estimate · Details
300 NVIDIA GeForce RTX 3080 Ti
12GB VRAM
$0.180/hr $0.180/hr 50% 1 Premium Estimate · Details
317 V100 FHHL
16GB VRAM
$0.190/hr $0.190/hr 47% 1 Premium Estimate · Details
317 Tesla V100
16GB VRAM
$0.190/hr $0.190/hr 47% 1 Premium Estimate · Details
383 V100 SXM2
16GB VRAM
$0.230/hr $0.230/hr 48% 1 Premium Estimate · Details
392 RTX 4090
24GB VRAM
$0.235/hr $0.350/hr 54% 3 Premium Estimate · Details
400 RTX 2000 Ada
16GB VRAM
$0.240/hr $0.240/hr 42% 1 Premium Estimate · Details
417 RTX A5000
24GB VRAM
$0.250/hr $0.490/hr 44% 5 Premium Estimate · Details
417 RTX A4500
20GB VRAM
$0.250/hr $0.250/hr 28% 1 Premium Estimate · Details
433 RTX 4000 Ada
20GB VRAM
$0.260/hr $0.260/hr 27% 1 Premium Estimate · Details
467 A100 PCIe
80GB VRAM
$0.280/hr $0.280/hr 43% 1 Premium Estimate · Details
583 T4
16GB VRAM
$0.350/hr $0.978/hr 60% 3 Frontier Estimate · Details
650 L4
24GB VRAM
$0.390/hr $2.81/hr 44% 3 Frontier Estimate · Details
667 A40
48GB VRAM
$0.400/hr $0.800/hr 50% 3 Frontier Estimate · Details
683 A30
24GB VRAM
$0.410/hr $0.410/hr 49% 1 Frontier Estimate · Details
700 T4G
16GB VRAM
$0.420/hr $1.10/hr 57% 1 Frontier Estimate · Details
733 RTX 4000 Ada SFF
20GB VRAM
$0.440/hr $0.440/hr 80% 1 Frontier Estimate · Details
750 RTX A6000
48GB VRAM
$0.450/hr $0.640/hr 27% 6 Frontier Estimate · Details
757 A10
24GB VRAM
$0.454/hr $2.40/hr - 3 Frontier Estimate · Details
767 NVIDIA GeForce RTX 3090 Ti
24GB VRAM
$0.460/hr $0.460/hr 70% 1 Frontier Estimate · Details
817 V100 SXM2 32GB
32GB VRAM
$0.490/hr $0.490/hr 49% 1 Frontier Estimate · Details
833 Quadro RTX 6000
24GB VRAM
$0.500/hr $0.500/hr - 1 Frontier Estimate · Details
833 NVIDIA GeForce RTX 4070 Ti
12GB VRAM
$0.500/hr $0.500/hr 80% 1 Frontier Estimate · Details
833 RTX 4080
16GB VRAM
$0.500/hr $0.500/hr 68% 1 Frontier Estimate · Details
833 RTX A2000
6GB VRAM
$0.500/hr $0.500/hr 88% 1 Frontier Estimate · Details
833 RTX PRO 6000 MaxQ
96GB VRAM
$0.500/hr $0.500/hr - 1 Frontier Estimate · Details
917 V100
16GB VRAM
$0.550/hr $3.06/hr - 2 Frontier Estimate · Details
983 RTX 5080
16GB VRAM
$0.590/hr $0.590/hr 49% 1 Frontier Estimate · Details
1000 P4
8GB VRAM
$0.600/hr $0.600/hr 60% 1 Frontier Estimate · Details
1200 A100 40GB
40GB VRAM
$0.720/hr $1.25/hr -59% 6 Frontier Estimate · Details
1250 RTX 6000 Ada
48GB VRAM
$0.750/hr $0.830/hr 48% 5 Frontier Estimate · Details
1333 A6000
48GB VRAM
$0.800/hr $0.800/hr - 1 Frontier Estimate · Details
1383 RTX 5000 Ada
32GB VRAM
$0.830/hr $0.830/hr 49% 1 Frontier Estimate · Details
1433 L40S
48GB VRAM
$0.860/hr $1.10/hr 70% 4 Frontier Estimate · Details
1483 RTX 5090
32GB VRAM
$0.890/hr $0.890/hr 40% 1 Frontier Estimate · Details
1583 L40
48GB VRAM
$0.950/hr $0.990/hr 47% 3 Frontier Estimate · Details
1677 A10G
24GB VRAM
$1.01/hr $1.83/hr 50% 1 Frontier Estimate · Details
1917 A100 80GB
80GB VRAM
$1.15/hr $1.49/hr 17% 9 Frontier Estimate · Details
2150 NVIDIA A100 40GB SXM
40GB VRAM
$1.29/hr $1.29/hr - 1 Frontier Estimate · Details
2304 H100 NVL
94GB VRAM
$1.38/hr $2.23/hr 23% 2 Frontier Estimate · Details
2317 A100 80GB
80GB VRAM
$1.39/hr $3.67/hr 41% 2 Frontier Estimate · Details
2433 P100
16GB VRAM
$1.46/hr $1.46/hr 62% 1 Frontier Estimate · Details
2483 GH200
96GB VRAM
$1.49/hr $1.49/hr - 1 Frontier Estimate · Details
2731 Gaudi $1.64/hr $1.64/hr 61% 1 Frontier Estimate · Details
3067 RTX PRO 6000
96GB VRAM
$1.84/hr $1.84/hr 35% 1 Frontier Estimate · Details
3317 MI300X
192GB VRAM
$1.99/hr $1.99/hr 25% 1 Frontier Estimate · Details
3483 RTX PRO 6000 WK
96GB VRAM
$2.09/hr $2.09/hr 44% 1 Frontier Estimate · Details
3500 H100 SXM
80GB VRAM
$2.10/hr $2.99/hr 17% 12 Frontier Estimate · Details
3833 H200
141GB VRAM
$2.30/hr $3.59/hr - 7 Frontier Estimate · Details
3983 H100 PCIe
80GB VRAM
$2.39/hr $2.49/hr 48% 3 Frontier Estimate · Details
5650 NVIDIA H200 NVL
143GB VRAM
$3.39/hr $3.39/hr - 1 Frontier Estimate · Details
5666 A100 40GB
320GB VRAM
$3.40/hr $3.75/hr 52% 2 Frontier Estimate · Details
6250 B200
192GB VRAM
$3.75/hr $3.99/hr 4% 3 Frontier Estimate · Details
8317 B200 SXM6
180GB VRAM
$4.99/hr $4.99/hr - 1 Frontier Estimate · Details
17667 H200 141GB
1128GB VRAM
$10.60/hr $10.60/hr - 1 Frontier Estimate · Details
20483 H100 80GB
640GB VRAM
$12.29/hr $12.29/hr 41% 1 Frontier Estimate · Details

How the GPU pricing index works

Index formula

index score = lowest tracked on-demand GPU-hour / cheapest tracked GPU-hour x 100

A score near 100 means the GPU is close to the cheapest tracked cloud rental baseline. A score of 400 means its lowest current GPU-hour is roughly four times that baseline.

Why median rate matters

The lowest listing can be limited by region, capacity, or marketplace host quality. Median on-demand pricing is a safer planning number when a team needs repeatable capacity.

Why provider count matters

A GPU with more providers is easier to source, compare, and negotiate. Single-provider bargains may be real, but they are less resilient for production budgets.

How to read each pricing tier

Value

Prototype and batch jobs

Start here for notebooks, image generation, small inference, and workloads where the model fits in cheaper memory.

Balanced

General AI production

Best for steady inference, fine-tuning, rendering, and ML jobs that need data-center reliability but not frontier memory.

Premium

High-memory workloads

Use when H100, H200, MI300X, or similar GPUs reduce sharding, speed up training, or make larger context windows practical.

Frontier

Blackwell and cluster capacity

Reserve this tier for workloads that clearly benefit from the newest accelerators, large nodes, or scarce enterprise capacity.

Provider benchmark shortcuts

RTX A4000

$0.060

Lowest tracked provider: TensorDock

Review all 4 provider rows

RTX 3090

$0.081

Lowest tracked provider: Vast.ai

Review all 3 provider rows

RTX 3070

$0.130

Lowest tracked provider: RunPod

Review all 1 provider rows

V100

$0.140

Lowest tracked provider: Datacrunch (Verda)

Review all 5 provider rows

NVIDIA GeForce RTX 3080

$0.170

Lowest tracked provider: RunPod

Review all 1 provider rows

NVIDIA GeForce RTX 3080 Ti

$0.180

Lowest tracked provider: RunPod

Review all 1 provider rows

V100 FHHL

$0.190

Lowest tracked provider: RunPod

Review all 1 provider rows

Tesla V100

$0.190

Lowest tracked provider: RunPod

Review all 1 provider rows

Common GPU pricing index decisions

Choosing between H100, H200, and B200

Start with the index score, then check whether memory bandwidth, VRAM, or Blackwell-specific performance reduces total runtime enough to justify the higher hourly rate.

Useful next steps: H100 calculator, H200 calculator, and B200 calculator.

Budgeting a rental cluster

Use the lowest rate for first-pass screening, median rate for budget approval, and provider count for capacity risk. For production inference, add idle endpoints, storage, network egress, support, and failed-run overhead.

Model the actual schedule in the GPU rental cost calculator.

Reference pricing sources

This page normalizes the prices tracked inside GPU Cost. For live market checks, compare against public provider and market references such as Runpod pricing, Vast.ai pricing, SemiAnalysis GPU Pricing Index, and AIMultiple GPU rental price index. Provider pages can change faster than indexed data, so confirm final quotes before committing to long jobs or reserved capacity.

GPU Pricing Index FAQ

A GPU pricing index normalizes cloud GPU rental prices so different accelerators can be compared from a common baseline. On this page, score 100 equals the cheapest tracked on-demand GPU-hour, and higher scores mean higher hourly cost.

No. The cheapest listing may have limited capacity, weaker networking, a less suitable region, or marketplace reliability risk. Use lowest price for screening, then compare median rate, provider count, workload fit, and total completed-job cost.

Spot discounts are useful for checkpointed training, batch inference, and interruptible experiments. For demos, production endpoints, and time-sensitive jobs, compare on-demand or reserved pricing instead of assuming spot availability.

Provider count shows how many tracked providers currently expose a GPU family. More providers usually means better capacity resilience, more region choices, and more room to compare support, networking, and commitment discounts.

Use this page to shortlist GPUs and understand relative pricing. Use the GPU rental cost calculator when you know the hourly rate, GPU count, utilization, schedule, storage, network, and overhead assumptions for a specific project.