GPU Rental Cost Calculator (Cloud GPU Cost per Job and Month)
Estimate what a cloud GPU job and a month of jobs will cost from the hourly rate, number of GPUs, job hours, idle time, storage and egress. Free, and it runs in your browser.
Updated November 19, 2026. Not financial advice. This calculator is for education and rough estimates only. It relies on simplified models and your inputs, ignores taxes and many real-world costs, and past or modelled results do not guarantee future outcomes. Consult a licensed professional before making financial decisions.
GPU and rate
Choosing a preset fills in an example rate. You can edit it.
Billed but not working: setup, downloads, waiting.
Data downloaded out of the cloud.
Many GPU clouds charge 0; hyperscalers often do not.
Estimated cost
Per job = rate × GPUs × (job hours + idle hours) + egress GB × egress rate. Per month = jobs × per job + storage GB × storage rate. No taxes, discounts, commitments or minimum charges.
Example GPU rates
Example on-demand rates checked on 2026-11-19. These are not live prices and not a complete or best-price comparison. Rates change and vary by provider, region, tier and availability.
| GPU | VRAM | $/hour | Source |
|---|---|---|---|
| NVIDIA H100 SXM 80 GB | 80 GB | $2.69 | RunPod (community cloud), checked 2026-11-19 |
| NVIDIA A100 SXM 80 GB | 80 GB | $1.39 | RunPod (community cloud), checked 2026-11-19 |
| NVIDIA L40S 48 GB | 48 GB | $0.79 | RunPod (community cloud), checked 2026-11-19 |
| NVIDIA RTX 4090 24 GB | 24 GB | $0.34 | RunPod (community cloud), checked 2026-11-19 |
Selected: NVIDIA H100 SXM 80 GB. Confirm the current rate on the provider's page before renting.
How to use the GPU rental cost calculator
- Pick a GPU to fill an example rate, or choose Other and type your provider's rate.
- Enter how many GPUs you need (check memory with the LLM VRAM calculator), the working hours per job, idle hours and jobs per month.
- Add storage and egress. Read the cost per job and per month.
- Compare with an API using the self-host vs API calculator, and estimate job length with the tokens per second calculator.
Worked example
One H100 at $2.69 an hour, 20 jobs a month of 8 working hours plus 1 idle hour each: 9 billed hours × $2.69 = $24.21 per job for the GPU. With 10 GB egress at $0 that stays $24.21. Over 20 jobs the GPU costs $484.20, and 200 GB of storage at $0.07 adds $14.00, for $498.20 a month. The idle hour alone costs $53.80 of that. The rates are examples, not quotes.
See also
- Before renting, work out how much VRAM you need for your model with the LLM RAM calculator.
- Estimate your token volume with the AI token counter.
Frequently Asked Questions (FAQ)
It depends on the GPU, provider and how many hours you use. As a rough guide, multiply the hourly rate by the hours you keep it running: at $1.00 per hour, a GPU running 24/7 for a 730-hour month costs about $730. Many providers bill per second, so a GPU used 40 hours costs only 40 hours. Enter your own numbers above.
No. The presets are example on-demand rates from one provider's public pricing page, labelled with the date they were checked and a link to the source. They are not live and not a complete market comparison. Prices vary by provider, region, tier, commitment and availability, so every rate is editable. Check the provider before you rent.
Idle time is the time the GPU is billed but not doing useful work, such as setup, downloading a model, waiting between runs or forgetting to stop an instance. Billed idle hours cost the same as working hours, so a job that works for 4 hours but sits for 2 more really costs 6 hours.
Often. Persistent disk or network volumes are usually billed per GB per month even when the GPU is stopped. Data leaving the provider (egress) is free at some GPU clouds and charged per GB at others, particularly the large hyperscalers. The defaults here are examples only; set them to your provider's rates.
Enough total VRAM for the model weights plus working memory. The LLM VRAM calculator on this site estimates this. Then enter that GPU count here; the cost scales with the number of GPUs.
For occasional or bursty work, renting usually is, because you pay only for hours used. For steady, high-volume workloads owning or committing can win. The self-host vs API calculator compares a rented or owned GPU against an LLM API.
Related Tools
- LLM API Cost CalculatorCompare GPT, Claude and Gemini API rates and estimate monthly token costs.
- RAG Chunking and Embedding Cost CalculatorEstimate RAG chunk count, one-off embedding (indexing) cost and monthly query cost from document size, chunk size, overlap, embedding price and retrieval settings. Prices are dated and editable.
- AI Token CounterCount tokens, words and characters in your prompts for GPT, Claude, Gemini and Llama.
- Tokens to Words, Pages and Books ConverterHow many tokens in a word, page or 1,000 words? Convert words, pages and books to tokens and back for six languages.
- LLM VRAM CalculatorEstimate how much GPU memory (VRAM) you need to run a local LLM, and which GPUs can run it.
- KV Cache CalculatorMemory per token and total KV cache from layers, KV heads, head dimension, context, batch and precision, with GQA and context-length views.