Should I rent a cloud GPU or buy one?
Renting is a running cost and buying is a fixed one, so the comparison is not a price but a threshold: the hours per month above which owning wins.
The short answer
Computed from this tool’s default settings — your hardware and the rest as most people start. Change them below for your own case.
At 40 hours a month, RTX 4090 pays back in about 63.8 months. Below roughly 213 hours a month, renting stays cheaper for a year or more.
Cloud GPU vs Buy Calculator
The hours per month at which renting stops being the cheaper option.
At 40 hours a month, RTX 4090 pays back in about 63.8 months. Below roughly 213 hours a month, renting stays cheaper for a year or more.
| Hours/month | Rent | Power if owned | Cheaper |
|---|---|---|---|
| 10 | $7 | $0.63 | Renting |
| 40 | $28 | $2.52 | Renting |
| 100 | $69 | $6.30 | Renting |
| 200 | $138 | $12.60 | Renting |
| 400 | $276 | $25.20 | Buying wins in year 1 |
| 730 | $504 | $45.99 | Buying wins in year 1 |
What each setting changes
Every input moves the result for a reason. This is what each one does and where to find the value for your own machine.
| Setting | Default | What it changes |
|---|---|---|
| Your hardware | RTX 4090 · 24 GB | The machine the model runs on. Usable memory decides what fits and memory bandwidth decides how fast it runs, so this moves every figure below. |
| Cloud hourly rate | 0.69 per hour | Community clouds run roughly $0.30–$0.80/hr for a 24GB card. |
| Hours per month | 40 hours | Anywhere from 1 to 730 hours. |
| Power draw under load | 350 W | Board power while generating, not idle. Card TDP is a good starting point. |
| Electricity price | 0.18 per kWh | Your actual tariff. US average is near $0.17, UK near £0.25, India near ₹8. |
Real answers across your hardware
The same calculation run at a range of settings, with everything else left at its default. These are computed by the tool itself, not written by hand.
| Your hardware | Buying wins after | Rent per month | Own per month | Card price |
|---|---|---|---|---|
| NVIDIA B200 · 180 GB | 0.0 months | $27.60 | $2.52 | n/a |
| Cerebras WSE-3 · 44 GB on-chip SRAM | 0.0 months | $27.60 | $2.52 | n/a |
| Raspberry Pi 5 · 16 GB | 4.8 months | $27.60 | $2.52 | $120 |
| RTX 4080 SUPER · 16 GB | 39.8 months | $27.60 | $2.52 | $999 |
| RTX 5070 · 12 GB | 21.9 months | $27.60 | $2.52 | $549 |
At 40 hours a month, NVIDIA B200 pays back in about 0.0 months. Below roughly 0 hours a month, renting stays cheaper for a year or more.
How this is calculated
No lookup tables and no invented constants. Here is the arithmetic, so you can check it against your own numbers.
Renting is a pure running cost; buying is a fixed cost plus electricity. Break-even is where the fixed cost, spread over the hours you actually use, drops below the rental rate minus your power cost.
Renting also avoids the thing this arithmetic cannot price: a card you own is three years older in three years, while a rented one is whatever the provider racked last quarter.
These are well-founded engineering estimates, not benchmark results. Your quantisation, runtime and context length all move the real number, and usable memory is an assumption rather than a specification. See the full methodology for every assumption behind these figures.
Who this is for
The three situations that bring people to this calculation.
Occasional use
Confirm renting is cheaper for a few hours a week.
Steady workloads
Find the hours where buying takes over.
Fine-tuning
Price a one-off training run against a purchase.
How to use this calculator
Four steps, no account, nothing leaves your browser.
Set your hardware
Start at the top of the panel. Every figure recalculates as you change it — there is no submit button, because watching the number move is the point.
Adjust the rest to match your setup
4 further settings: cloud hourly rate, hours per month, power draw under load, electricity price. Defaults are the common case, so change only what differs for you.
Read the headline, then the table
The large figure answers the question. The table underneath shows how the answer changes across nearby settings, which is usually where the decision actually gets made.
Check it against the method
The arithmetic is written out above. If a number looks wrong for your hardware, the assumptions are the first place to look — usable memory and quantisation are the two that vary most.
Should I rent a cloud GPU or buy one: common questions
The questions people ask about this, answered without hedging.
Is renting a cloud GPU cheaper than buying?
For light and occasional use, almost always. Community clouds rent 24GB cards for well under a dollar an hour, and a few hours a week takes years to reach a card’s price.
When does buying win?
Once you are running most of the day, most days. At that point the fixed cost spreads thin and the rental meter does not stop.
What does renting have that buying does not?
No obsolescence and no commitment. A card you own is three years older in three years; a rented one is whatever the provider racked most recently. You can also rent hardware you would never buy.
What are the hidden costs of renting?
Storage between sessions, egress, and the time spent re-uploading models and data on every cold start. For iterative work that friction is a real cost.
Related local AI tools
All 50 run on the same arithmetic, so answers across them agree.
More on cost
Will it fit
Context
Speed
Quantisation
Runtime & errors
Fine-tuning
Searches this page answers
Different ways of asking the same question, all resolved above.
Now find the models that fit
Sizing is only half the problem. Model Radar takes your hardware and shows which models actually run on it, ranked by what they are good at — the same arithmetic as this page, applied to every model worth running.