Price checkPer GPU
Community Cloud (Spot)CustomSecure Cloud (On-Demand)CustomServerlessCustom
See all 3 plans
Quick Answer
Last verified:
High confidence

RunPod costs $0.34 to $3.49 per GPU/hour as of May 2026, with 3 plans available. The median contract is $400/year based on 15 verified purchases.

Use the interactive pricing calculator to estimate your exact cost based on team size and requirements.

  • Free tier: No free tier available

RunPod offers 3 pricing tiers: Community Cloud (Spot), Secure Cloud (On-Demand), Serverless. The Secure Cloud (On-Demand) plan is production inference and training requiring reliable gpu availability.

Compared to other ai/gpu cloud compute software, RunPod is positioned at the budget-friendly price point.

  • Median contract: $400/yr from 15 purchases
  • 8 documented hidden costs beyond list price
Last verified: | Sources: 2 pricing pages and reviews

How much does RunPod cost?

RunPod pricing ranges from $0.34 to $3.49/GPU/hour across 3 plans. Plans include Community Cloud (Spot) (custom pricing), Secure Cloud (On-Demand) (custom pricing), Serverless (custom pricing).

RunPod Pricing Overview

RunPod has 3 pricing plans ranging from $0.34 to $3.49/GPU/hour. The Community Cloud (Spot) plan requires contacting sales for a custom quote and is designed for batch workloads, training runs, and cost-sensitive inference that can tolerate interruptions. The Secure Cloud (On-Demand) plan requires contacting sales for a custom quote and is designed for production inference and training requiring reliable gpu availability. The Serverless plan requires contacting sales for a custom quote and is designed for low-traffic endpoints and batch api serving with infrequent requests.

The median RunPod customer pays $400/year based on 15 verified purchases.

There are at least 8 documented hidden costs beyond RunPod's list price, including implementation, training, and add-on fees.

This pricing was last verified in April 24, 2026 from 2 independent sources.

Start RunPod Free Trial

RunPod is a pay-as-you-go GPU cloud platform with three compute tiers: Community Cloud (Spot) for budget-conscious batch jobs starting around $0.22/hr for an RTX 3090 and up to $3.49/hr for an H100 SXM on-demand, Secure Cloud (On-Demand) for guaranteed availability at a premium, and Serverless for per-request workloads billed by active compute time. There are no monthly subscriptions — you prepay credits and pay only for what you use, billed by the second with no minimum commitment required.

How RunPod Pricing Compares

Compare RunPod pricing against top alternatives in AI/GPU Cloud Compute.

Amazon Web Services (AWS) $0-$10000/month Compare → Microsoft Azure $1.0-$12.0/month Compare → DigitalOcean $4-$200/month Compare →

All RunPod Plans & Pricing

Plan Monthly Annual Best For
View all features by plan (compare side-by-side)

Community Cloud (Spot)

  • Consumer GPUs from third-party hosts
  • RTX 3090: ~$0.22/hr
  • RTX 4090: ~$0.34/hr
  • ~50% cheaper than Secure Cloud
  • Preemptible — instances may be reclaimed

Secure Cloud (On-Demand)

  • Datacenter-grade GPUs
  • RTX 4090: ~$0.69/hr
  • A100 80GB: ~$1.99/hr
  • H100 SXM: $3.49/hr
  • H100 PCIe: $2.99/hr
  • Persistent instances with guaranteed uptime

Serverless

  • Auto-scaling endpoints
  • Pay only for active inference time
  • Zero idle cost
  • Cold starts: 30–60 seconds
  • Custom model deployment
Start RunPod Free Trial
Compare RunPod with alternativesAdjust seats, lock a tier, add up to 2 more products side-by-side. Shareable URL.

Usage-Based Rates

Per-unit pricing for RunPod API usage.

Community Cloud (Spot)

Model Unit Rate
RTX 3090 (24GB) second $0.000061 Community Cloud spot (~$0.22/hr)
RTX 4090 (24GB) second $0.000094 Community Cloud spot (~$0.34/hr)
A100 80GB second $0.00046 Community Cloud spot (~$1.64/hr)
H100 SXM (80GB) second $0.00075 Community Cloud spot (~$2.69/hr)
  • Community Cloud prices are approximate — vary by availability
  • Instances may be reclaimed; use for interruptible workloads only
  • Price expressed per second; multiply by 3600 for hourly equivalent

Secure Cloud (On-Demand)

Model Unit Rate
RTX 4090 (24GB) second $0.00019 Secure Cloud on-demand (~$0.69/hr)
A100 80GB second $0.00055 Secure Cloud on-demand (~$1.99/hr)
H100 PCIe (80GB) second $0.00083 Secure Cloud on-demand (~$2.99/hr)
H100 SXM (80GB) second $0.00097 Secure Cloud on-demand (~$3.49/hr)
  • 3-month reserved commitment reduces H100 from $3.49/hr to $2.79/hr (~20% off)
  • Network volume storage: $0.07/GB/month (billed even when pod is stopped)
  • Price expressed per second; multiply by 3600 for hourly equivalent

Serverless

Model Unit Rate
Serverless (RTX 4090 worker) second $0.00020 Approximate active GPU-second rate
Serverless (A100 worker) second $0.00056 Approximate active GPU-second rate
  • Cold start time (30–60s) is billed — use keep-alive workers for latency-sensitive endpoints
  • Exact serverless GPU rates depend on worker configuration

Compare RunPod vs Alternatives

Before committing to RunPod, compare pricing with these 3 alternatives in the same category.

VSAmazon Web Services (AWS)
Free

Developers learning AWS, small experiments, and proof-of-concept projects

Full comparison
VSMicrosoft Azure
From $1.0/month

Developers and startups exploring Azure or building initial prototypes

Full comparison
VSDigitalOcean
From $4/month

Evaluating DigitalOcean services

Full comparison

All RunPod alternatives & migration guides

What Companies Actually Pay for RunPod

The median RunPod buyer pays $400/year based on 15 verified purchase transactions.

What companies actually pay $400/yr Median across 15 community cost mentions
Review scores
Top pricing complaints
Serverless cold start times make real-time or low-latency applications impracticalReliability issues with Community Cloud and tier 2/3 data center GPU poolsNetwork volume storage costs accumulate even when pods are stoppedComplex manual setup and configuration required before paying-by-the-hour compute begins
Source: Community cost mentions (Reddit, Hacker News) — aggregated from 15 distinct user reports. Indicative only — not contract-grade data.

RunPod Year 1 Total Cost by Company Size

Real deployment costs including licenses, implementation, training, and admin — not just the sticker price.

Whisper Transcription - 400 Hours of Audio $30 Year 1 total

Transcribing 400 hours of audio using Whisper Large v2 on RunPod RTX A5000 or RTX 3,090 GPUs at $0.22/hour

Daily AI Inference Workload $262 Year 1 total

Running A100 80GB for inference at $1.64/hour for 8 hours per day over 20 working days

Video Generation with Hunyuan $96 Year 1 total

6 hours of H100 GPU time for training e-girl/influencer LoRA model for Hunyuan video at $2.69-$2.99/hour per GPU

12-Hour Training Run on Community Cloud ~$10 for 12 hours on Community Cloud Year 1 total

A single overnight training or fine-tuning run on a Community Cloud (Spot) GPU instance — typical for LoRA fine-tunes or short ML experiments.

Always-On 13B LLM Inference Server ~$1,500/month estimated for always-on deployment Year 1 total

Hosting a 13B parameter language model on a RunPod pod 24/7 for production RAG or API inference.

Developer Workday GPU Rental (8 hrs/day) ~$415/month ($2.59/hr × 8 hrs × 20 days) Year 1 total
$2.59/hr × 8 hrs × 20 days
Total ~$415/month ($2.59/hr × 8 hrs × 20 days)

Renting a single GPU for standard working hours (8 hours/day, 20 working days/month) for regular ML development or model testing. Based on an H100-class GPU at approximately $2.59/hr.

Short Burst Creative or Rendering Job ~$1 for a 288-frame animation (RTX 4090 at $0.69/hr) Year 1 total
RTX 4090 at $0.69/hr
Total ~$1 for a 288-frame animation (RTX 4090 at $0.69/hr)

Using a RunPod RTX 4090 for a brief rendering or image generation job lasting well under one hour.

100-Hour ML Training Run (RTX 3080) ~$17 for 100 hours on RTX 3080 Year 1 total

A 100-hour training run on a RunPod RTX 3080 instance — comparable to training a small transformer model from scratch.

Reddit discussion comparing transcription costs across providers

How RunPod Pricing Compares

Software Starting Price Top Price
RunPod $0.34/GPU/hour $3.49/GPU/hour
CoreWeave $10/instance/hour $68.8/instance/hour
Hyperbolic $0.3/GPU/hour $3.2/GPU/hour
Lambda $0.69/GPU/hour $6.99/GPU/hour
Modal Free $250/GPU/hour
Paperspace Free $39/GPU/hour

Detailed pricing comparisons:

Browse all AI/GPU Cloud Compute pricing →

8 RunPod Hidden Costs Beyond the List Price

Beyond the listed price, RunPod has at least 8 documented hidden costs that can significantly increase total cost of ownership.

Watch for 8 hidden costs
  • Network Volume Storage Fees When Pod Is Stopped $0.07-$0.20/GB/month
    medium 2 sources
    Reddit "when you stop the GPU, you still need to pay for the storage which is way more than a normal cloud storage pricing (min 0.07$/GB/month for RunPod using Network Volume)"
    Reddit "2 A40s will cost (GPU $0.78/hr + Pod $0.017/hr) while running and $0.028/hr while stopped"
  • Container Startup Time Billing $0.50-$2 per startup
    low 1 source
    Reddit "No matter how much or how little you use it, you'll pay for the time it's running (typically, including startup time, which makes docker images that have huge dependency trees or otherwise take a long time to start up fresh a bit more expensive)."
  • Serverless Cold Start and Keepalive Idle Costs 20-40% of compute costs
    high 3 sources
    Hacker News "I spent a month setting up serverless endpoint for a custom model last year with Runpod. It was expensive and unreliable, in addition to long cold boot times."
    Reddit "the longer the keepalive, the more you pay while it is doing nothing. They're really great for small custom models that load in a second, but you will pay for the entire VRAM of the desired GPU target."
    Reddit "For Runpod, you're looking at their on-demand pricing. Their serverless H100s are $5.59/hr (scroll to bottom of https://www.runpod.io/serverless-gpu)."
  • Community Cloud Reliability Issues Lost work time + redeployment costs
    high 2 sources
    Reddit "RunPod and Vast are great, but we've heard from our users that they're prone to reliability issues since they're built on mostly lower tier providers and community GPU pools."
    Reddit "there have been countless reports and stories of people losing their GPU when using Runpod"
  • Serverless Markup Over On-Demand Pod Pricing 20-160x on-demand pod rate
    high 1 source
    Hacker News "if you use the 'serverless' services, the pricing becomes even more astronomical; as you note, $1/minute is unreasonably expensive: that's over 20x the cost of renting 8xH100s on Runpod's 'Secure Cloud' (and 8xH100s are extreme overkill for finetu..."
  • Community Cloud Preemption Risk and Re-Setup Costs 10-30% of compute costs for interrupted jobs
    medium 2 sources
    Reddit "the 4090s are 0.69/hour for the 'secure cloud' and starting at 0.34/hour for the 'community cloud'. I thought this meant if you pay 0."
    Reddit "RunPod (or similar): You can install anything you want, but you have to install it yourself, you have to pay for the UI uptime even when the GPU is doing nothing, you have to pay for storage when the pod is stopped, and you have to hope that when ..."
  • Startup and Docker Provisioning Time Billing $0.01-$0.10/session
    low 2 sources
    Reddit "No matter how much or how little you use it, you'll pay for the time it's running (typically, including startup time, which makes docker images that have huge dependency trees or otherwise take a long time to start up fresh a bit more expensive)."
    Reddit "Runpod is good pricing but I hate spending forever trying to figure how to get an app installed while in paying by the hour."
  • Production Reliability Risk from Tier 2/3 Data Centers 5-20% of license costs
    high 3 sources
    Reddit "Runpod isn't reliable for production because they use tier 2 and 3 data centers which is also what allows them to offer such low prices. So you get better prices but with lower stability."
    Reddit "I tried runpod and it isnt reliable for production. I tried to generate image with 24gb gpu, and request completion time varies a lot."
    Hacker News "I spent a month setting up serverless endpoint for a custom model last year with Runpod. It was expensive and unreliable, in addition to long cold boot times."
Tip

Ask your RunPod sales rep about these costs upfront. Getting them in writing before signing can save you from surprise charges later.

Full hidden costs breakdown →

Intelligence sourced from 2 independent sources
Reddit User discussions Hacker News Tech community
Key claims include inline source attribution. Data verified against multiple independent sources. 29 source citations total.

RunPod Contract Terms

RunPod contracts do not auto-renew. Changes require No contract — stop or delete pods at any time, no notice required. These terms are sourced from verified buyer experiences.

Contract Terms
Auto-Renewal No
Cancellation Notice No contract — stop or delete pods at any time, no notice required
Minimum Commitment None for on-demand pods; 3-month commitment available for discounted H100 pricing
Mid-Term Downgrade Allowed
Payment Terms Prepaid credits, pay-as-you-go billed per second of active pod usage
Price Escalation No published uplift schedule; market pricing has generally trended downward with increased competition
Note

Users can switch GPU types, tiers, or delete pods freely at any time

Based on 2 verified sources

How to Negotiate RunPod Pricing

RunPod contracts are negotiable. These 10 tactics are sourced from real buyer experiences and procurement specialists.

Negotiation Playbook 10 tactics
Use Community Cloud for 50% Savings high success

Community Cloud instances cost approximately 50% less than Secure Cloud (e.g., RTX 4,090 at $0.34/hr vs $0.69/hr), but with the tradeoff of potential instance preemption.

Reddit user reporting Community Cloud pricing at half of Secure Cloud rates
Commit to 3-Month Reserved Capacity high success

Committing to 3 months reduces H100 pricing from $3.49/hr to $2.79/hr (20% discount). This is RunPod's longest commitment period.

Hacker News comment comparing RunPod reserved vs on-demand pricing
Use Container Disk Instead of Network Volumes high success

Avoid network volumes ($0.07/GB/month minimum) by using larger container disk allocations, which only cost during active usage. Delete network volumes immediately when no longer needed.

Reddit advice on avoiding unnecessary network volume charges
Use Community Cloud (Spot) for Non-Critical Workloads high success

Community Cloud (Spot) instances cost roughly half the price of Secure Cloud (On-Demand) for the same GPU. An RTX 4090 runs about $0.34/hr on Community Cloud vs $0.69/hr on Secure Cloud; an A100 is ~$1.19/hr vs ~$1.89/hr. For batch processing, one-off training runs, fine-tuning experiments, or any workload that can tolerate interruptions and restart from a checkpoint, using Community Cloud is the single most effective cost lever.

reddit community pricing comparisons
Commit 3 Months on H100 to Lock In ~20% Discount high success

RunPod's longest commitment tier reduces H100 SXM pricing from $3.49/hr (on-demand) to $2.79/hr — approximately a 20% savings. If you have consistent H100 demand for training or inference over a 3-month period, this commitment pays for itself in a matter of days. At 8 hours/day use, the savings exceed $400/month on a single H100.

HN (reissbaker, 2024-09-12)
Delete Network Volumes Immediately When Not Needed high success

Persistent network volumes charge $0.07/GB/month minimum even while pods are stopped. For large model collections (50–200GB), this means $3.50–$14/month in idle storage costs. The workaround is to save model weights to Hugging Face Hub or object storage between sessions and re-download at pod startup — adding a few minutes of startup time but eliminating ongoing storage charges entirely.

reddit (StableDiffusion and LocalLLaMA subreddits)
Use On-Demand Pods Over Serverless for Sustained Workloads high success

RunPod Serverless pricing can be 20x or more expensive per GPU-hour than an equivalent on-demand pod for sustained workloads. Serverless H100s list at $5.59/hr vs $2.69–$2.99/hr for on-demand H100 pods. For any workload with predictable or moderate traffic, renting a pod and managing uptime manually delivers dramatically better economics than paying the serverless markup.

HN and reddit (MachineLearning subreddit)
Use Spot Pricing for Interruptible Workloads medium success

Spot pricing offers 50% discount over on-demand rates, suitable for training jobs that can tolerate interruptions.

Reddit discussion of RunPod Spot vs on-demand pricing
Benchmark on Cheaper GPUs Before Committing to High-VRAM Options medium success

RunPod's GPU catalog spans a wide price-performance range. A pair of A40s at $0.78/hr for both may handle the same inference workload as a single H100 at $2.69+/hr. Testing your specific workload on cheaper options first — particularly for inference where throughput rather than raw VRAM is the bottleneck — can reveal significant savings before committing to an expensive GPU for a long training run.

reddit community benchmarking discussions
Compare Marketplace Competitors Before Renting medium success

Multiple users highlight that Vast.ai, Shadeform, Lambda Labs, and TensorDock regularly undercut RunPod on specific GPU types. A100 80GB starts at $0.67/hr on Vast.ai vs $1.19/hr on RunPod Community Cloud; H100s can be found for $1.90/hr on Shadeform vs $2.69/hr on RunPod. Always check competing marketplaces for the specific GPU type and region you need before committing.

reddit pricing comparison table (LocalLLaMA subreddit, 2025-03-06)
High success Medium Low / difficult

Full negotiation guide →

RunPod Pricing FAQ

01 What's the difference between Secure Cloud and Community Cloud?

Secure Cloud uses datacenter GPUs at standard rates (e.g., RTX 4,090 at $0.69/hr), while Community Cloud uses consumer GPUs from third parties at ~50% discount (e.g., RTX 4,090 at $0.34/hr). Community Cloud instances can be preempted without notice, making them less reliable for production workloads.

02 Do I pay for storage when my GPU is stopped?

Yes, network volumes continue billing at $0.07/GB/month minimum even when your pod is stopped. Container disk storage only charges during active pod runtime. To avoid ongoing storage costs, delete network volumes when not needed or use larger container disk allocations instead.

03 Is RunPod serverless cost-effective?

RunPod serverless has significant cold start times (30+ seconds) and can be expensive compared to on-demand pods. One user reported it was 'expensive and unreliable' for custom model hosting. For sustained usage, hourly pod rental is more cost-effective than serverless.

04 How does RunPod pricing compare to AWS/GCP?

RunPod is significantly cheaper than major cloud providers. H100s cost $2.69-$3.49/hr on RunPod vs much higher on AWS/GCP. However, RunPod uses tier 2-3 datacenters, trading some reliability for lower costs compared to hyperscaler infrastructure.

05 What is the difference between Community Cloud (Spot) and Secure Cloud (On-Demand) on RunPod?

Community Cloud (Spot) uses third-party and community-contributed GPUs, making it roughly 50% cheaper — for example, an RTX 4090 is about $0.34/hr on Community Cloud vs $0.69/hr on Secure Cloud. The tradeoff is that Community Cloud instances can be preempted at any time without warning, forcing a restart and potential loss of unsaved progress. Secure Cloud (On-Demand) uses RunPod's own data center infrastructure and provides more reliable uptime. Use Community Cloud for batch jobs and experiments; use Secure Cloud for anything requiring consistent availability.

06 Does RunPod charge when my pod is stopped?

Yes — if you have a network volume (persistent storage) attached, you continue paying storage fees even when the pod is not running. RunPod's minimum network volume pricing starts at $0.07/GB/month. A pod with 2x A40 GPUs costs approximately $0.028/hr while stopped (storage only) versus $0.78/hr while running. To avoid idle storage costs, delete your network volume when not needed and re-download models on pod startup from Hugging Face Hub or object storage.

07 Is RunPod reliable enough for production AI workloads?

RunPod uses tier 2 and tier 3 data centers alongside community GPU pools to keep pricing low — this means production reliability is inconsistent. Multiple users report that RunPod Serverless endpoints have variable cold start times and unpredictable request completion times, making it unsuitable for latency-sensitive production workloads. Secure Cloud (On-Demand) pods are more stable than Community Cloud but still fall short of enterprise-grade SLAs. For production use, users often recommend Lambda Labs, Crusoe, or hyperscaler alternatives, accepting a higher per-GPU-hour cost.

08 How does RunPod Serverless pricing compare to renting an on-demand pod?

RunPod Serverless is significantly more expensive per GPU-hour than an equivalent on-demand pod. Serverless H100s are listed at $5.59/hr versus $2.69–$2.99/hr for on-demand H100 pods — roughly a 2x markup. For burst workloads like image generation at scale, one HN commenter noted serverless pricing can represent 20–160x the equivalent pod cost depending on GPU configuration. For sustained or predictable workloads, on-demand pods offer dramatically better economics; serverless is only advantageous for very sporadic, low-frequency request patterns.

09 What are the cheapest GPU options on RunPod?

As of early 2025, the cheapest GPUs on RunPod include the RTX 3080 (10GB) starting around $0.17/hr, RTX A4000 (16GB) from $0.17/hr, RTX 3090 (24GB) from $0.22/hr on Community Cloud, A40 (48GB) at $0.44/hr, and A6000 (48GB) from $0.44/hr on Community Cloud. All prices are for Community Cloud (Spot); Secure Cloud rates are higher. Prices vary by availability and fluctuate with demand.

10 Can I get discounted pricing with a longer commitment on RunPod?

Yes. RunPod offers committed pricing on H100 instances — a 3-month commitment reduces the H100 SXM on-demand rate from $3.49/hr to $2.79/hr, approximately a 20% discount. This is the longest commitment period RunPod currently offers. For all other GPU types and shorter workloads, pay-as-you-go remains the standard and most flexible option.

Is this pricing incorrect? — we'll verify and update it.