Pricing & Commercial
RunPod prices each product differently. Community Cloud follows marketplace dynamics; Secure Cloud follows operator economics; Serverless follows function-call pricing. Per-second billing across all three.
Community Cloud pricing
Community Cloud prices are set by individual providers within RunPod's marketplace structure. The platform suggests pricing but doesn't enforce specific levels.
Indicative pricing (illustrative, varies):
| GPU | Community Cloud | Vast equivalent |
|---|---|---|
| RTX 4090 | $0.35-0.55/hr | $0.30-0.45/hr |
| A100 80GB | $1.20-1.70/hr | $1.00-1.40/hr |
| H100 80GB | $2.50-3.50/hr | $2.00-3.00/hr |
| H200 | $3.00-4.00/hr | $2.50-3.50/hr |
The 10-25% premium to Vast reflects the better DX, more curated supply, and platform integration. For most users, the premium is worth it; for hardcore price optimizers, Vast wins.
Secure Cloud pricing
Secure Cloud is RunPod-priced, not marketplace-determined. Pricing is typically 20-50% above Community Cloud for equivalent hardware, reflecting the dedicated reliability tier.
| GPU | Secure Cloud | Hyperscaler list | Discount to hyperscaler |
|---|---|---|---|
| A100 80GB | $1.70-2.20/hr | $4-5/hr | ~55% |
| H100 80GB | $3.50-4.50/hr | $8-12/hr | ~55% |
| H200 | $4.50-5.50/hr | $10-14/hr | ~55% |
Secure Cloud is roughly half the price of hyperscaler list — competitive with Lambda Cloud and other mid-tier enterprise neoclouds. Not as cheap as the marketplace but positioned for customers who need consistency.
Serverless pricing
Serverless prices per-second of GPU compute, billed only while the function is running. No idle cost when no requests are arriving.
The pricing makes serverless attractive for low-utilization inference workloads. If you receive 1000 requests per day and each takes 2 seconds of GPU time, you pay for ~2000 GPU-seconds total, not 86400 seconds of dedicated GPU. The savings can be 50x+ for sparse workloads.
For high-utilization workloads, dedicated pod or Secure Cloud is cheaper on a per-hour basis.
Per-second billing
RunPod bills by the second (with some minimum). This is a meaningful improvement over the hourly billing many smaller clouds use.
Impact:
- Short jobs (a few minutes of inference, brief experimentation) don't pay for a full hour.
- Restarted instances after preemption don't lose partial hours.
- Per-second billing matches per-second cost; the user's bill tracks actual usage.
For workloads that spin up and tear down frequently, this can be material.
Reservations & commitments
RunPod offers reserved pricing on Secure Cloud for customers willing to commit. The discount versus on-demand grows with the commitment term:
- Monthly commitment: ~10-15% discount.
- Quarterly commitment: ~20-25% discount.
- Annual commitment: ~30-40% discount.
This is the pricing structure that competes with CoreWeave / Crusoe / Lambda for enterprise customers. The discounts narrow the per-hour cost gap to bring RunPod competitive for committed workloads.
Contracts & enterprise terms
For larger customers, RunPod offers enterprise contracts with:
- Master service agreements (MSAs) with negotiated terms.
- Custom pricing on volume.
- Account management and prioritized support.
- SOC 2 attestation as the contracting baseline.
- Optional dedicated capacity reservations.
The enterprise motion is less mature than at CoreWeave or Crusoe — RunPod's headcount and sales infrastructure are smaller. But the offering exists and lands deals at meaningful annual contract values.
Comparative position
RunPod's pricing position by competitor:
- vs Vast (marketplace): 10-25% more expensive on Community Cloud; better DX and somewhat better reliability.
- vs Lambda (mid-tier enterprise): Comparable on Secure Cloud; RunPod has the marketplace channel Lambda doesn't.
- vs CoreWeave (Platinum-tier enterprise): Cheaper on Secure Cloud for smaller commitments; CoreWeave wins on huge multi-year reservations with bespoke pricing.
- vs hyperscalers: 50-60% cheaper across the board; comparable to other mid-tier neoclouds.
- vs Together.AI / Fireworks (inference): Serverless competes head-on; pricing varies by model and workload.
Takeaway
RunPod's pricing reflects its blended business model. Cheap-enough on the marketplace side, competitive-enough on the dedicated side. Per-second billing is a meaningful customer-friendly feature. The next chapter covers the hardware behind the pricing.