How much is an H100 on RunPod?

H100 SXM lists at $2.69/h on Community Cloud and $3.99/h on Secure Cloud; H100 PCIe $1.99 / $2.89, H100 NVL $2.59 / $3.19, H200 $3.59 / $5.29, B200 $5.98 / $7.99, A100 SXM $1.39 / $1.79, L40S $0.79 / $1.09, RTX 4090 $0.34 / $0.89.

Community vs Secure Cloud — what's the difference?

Community Cloud runs on vetted third-party hosts and is cheaper; Secure Cloud runs in Tier 3/4 data centers with stronger reliability and compliance, at a higher hourly rate.

Do idle pods still cost money?

Yes. A running pod bills every wall-clock hour. At 20% utilization an H100 SXM community pod at $2.69/h costs $13.45 per useful GPU-hour.

What does RunPod Serverless cost?

Serverless flex workers bill only active time: B300 $11.99/h, B200 $10.79, H200 $6.99, H100 $5.39, A100 $2.72, L40/L40S/6000 Ada $1.75, RTX 4090 $1.34, L4/A5000/3090 $0.75. The H200 $6.35 and A100 $2.15 figures sometimes quoted are Cluster rates, not Serverless.

When is a pod cheaper than serverless?

When utilization exceeds pod rate ÷ serverless rate. For H100 SXM community: $2.69 ÷ $5.39 ≈ 50% — busier than that, keep a pod; spikier, use serverless.

Calculators

RunPod GPU Cost Calculator

Pick a RunPod GPU (B300 to RTX A5000), Community or Secure Cloud, GPUs per pod, hours per day, days per month and utilization. The calculator returns monthly pod cost, cost per useful GPU-hour (idle pods still bill) and the RunPod Serverless cost for the same busy hours with the break-even utilization. Estimates only; runpod.io/pricing prevails.

Math runs in your browser. Nothing is uploaded.

GPU

Cloud

GPUs per pod

Hours / day running

Days / month

Utilization % (busy share of running time)

Pod cost
—

Effective cost
—

Serverless alternative
—

Formula: pod cost = GPUs × hours/day × days × $/GPU-hour — a running pod bills wall-clock time whether busy or idle. Cost per useful GPU-hour = rate ÷ utilization. Serverless (flex workers) = useful hours × serverless $/h, so it wins below the break-even utilization pod rate ÷ serverless rate. Serverless rates per GPU class: B300 $11.99, B200 $10.79, H200 $6.99, H100 $5.39, RTX 6000 Pro $3.49, A100 $2.72, 5090 $1.79, L40/L40S/6000 Ada $1.75, 4090 $1.34, A6000/A40 $1.22, RTX PRO 4500 Blackwell $1.15, A4000-class (16 GB) $0.58, L4/A5000/3090 $0.75 per hour. Storage and network volumes are extra.

Estimates only. Official list prices as of 2026-11-01 (Asia/Shanghai); the provider's pricing page prevails. Sources: RunPod pricing

GPU Cloud Cost Comparison · Modal GPU Cost Calculator · fal.ai Cost Calculator · Together AI Token Cost Calculator · Fireworks Token Cost Calculator · DeepInfra Token Cost Calculator · Cloudflare Workers AI Cost Calculator

How it works

How much is an H100 on RunPod?: H100 SXM lists at $2.69/h on Community Cloud and $3.99/h on Secure Cloud; H100 PCIe $1.99 / $2.89, H100 NVL $2.59 / $3.19, H200 $3.59 / $5.29, B200 $5.98 / $7.99, A100 SXM $1.39 / $1.79, L40S $0.79 / $1.09, RTX 4090 $0.34 / $0.89.
Community vs Secure Cloud — what's the difference?: Community Cloud runs on vetted third-party hosts and is cheaper; Secure Cloud runs in Tier 3/4 data centers with stronger reliability and compliance, at a higher hourly rate.
Do idle pods still cost money?: Yes. A running pod bills every wall-clock hour. At 20% utilization an H100 SXM community pod at $2.69/h costs $13.45 per useful GPU-hour.
What does RunPod Serverless cost?: Serverless flex workers bill only active time: B300 $11.99/h, B200 $10.79, H200 $6.99, H100 $5.39, A100 $2.72, L40/L40S/6000 Ada $1.75, RTX 4090 $1.34, L4/A5000/3090 $0.75. The H200 $6.35 and A100 $2.15 figures sometimes quoted are Cluster rates, not Serverless.
When is a pod cheaper than serverless?: When utilization exceeds pod rate ÷ serverless rate. For H100 SXM community: $2.69 ÷ $5.39 ≈ 50% — busier than that, keep a pod; spikier, use serverless.

RunPod GPU Cost Calculator

More in this group

How it works