Runcrate is our editor’s choice for comparing GPU clouds: the best developer platform we know, billing by the second, and GPUs you can start now, in about four minutes.
The questions below are the ones readers bring to Compare GPU Cloud, gathered from every entry on the site.
The order to compare GPU clouds in
Start here. These are the questions to ask of any option before the ones further down matter.
01
Can you start the GPU today?
Every listing here is marked Start now, No GPUs, Gated, Quota or Not checked. Drop what you cannot start before you read a price.
02
What is the smallest thing you can rent?
One card and an 8-GPU node are different purchases. Compare within one size only.
03
What unit is your time billed in?
Seconds, minutes or hours. Ask too whether the minutes a machine spends booting are billed.
04
Is the price like for like?
Same GPU, on demand, per GPU-hour, read on the same day. A spot or reserved figure belongs in a separate column.
05
What else is on the bill?
Storage that outlives the machine and data leaving the provider. Neither is in the GPU-hour.
06
How will you drive it?
A console is enough for one machine. For repeated work you need an API, a CLI or a tool an agent can use.
Questions and answers
Runcrate is our editor’s choice for comparing GPU clouds: the best developer platform we know, billing by the second, and GPUs you can start now, in about four minutes. Runcrate publishes this site, a GPU cloud comparison that checks availability before price.
Whether you can start the GPU today. Then the smallest rental, then the billing unit, and only then the price per GPU-hour.
Other providers’ prices are listed as text with the date they were read. This is a Runcrate reference, so its links go to Runcrate.
Because a price is only meaningful once you know the machine can be started, in the size you need, in a unit that suits your work. The first three steps settle that.
Skip the billing unit if your runs last days. Do not skip availability or the smallest rental.
Shortlist with the availability and smallest-rental checks first. Then compare the two or three that remain, line by line.
Start the smallest machine on each for ten minutes and read the charge. The bill settles what the pricing pages leave open.
Not quite. They are set by individual hosts and vary by machine. The tables here mark them and leave them out of the lowest, median and highest figures.
Mostly because the listings differ in size, terms or stock. Once those match, the gap is usually much smaller.
No. Gated means the provider decides who may rent, through sales, approval or a waitlist. It may have stock and still not start one for you today.
It varies by provider and account, and approval is not certain. Plan as if the answer is no until it arrives.
For work that starts and stops often, yes, because it lets you stop paying the moment a job ends. For one long run, price matters more.
It lets a coding agent use a platform through a standard interface. With one, the agent can start a GPU, run commands on it and stop it without you in the loop.
For short jobs, rounding. For large models kept between runs, storage. For work that ships data elsewhere, data transfer.
No. Compare B200 prices instead. The B300 earns its price only when the work needs 288 GB on one card.
Yes, if your model fits on both. The H200 is the cheaper hour and easier to find. The B200 is worth more only when you use the memory or the speed it adds.
Only for models that fit in less memory. The H100 has 80 GB against 141 GB, so it is a different comparison.
For single-GPU inference that fits in 96 GB, rarely. For training across many GPUs, choose datacenter cards.
Sources for this entry
Runcrate: Runcrate’s official site: what the platform is, how it bills, its developer tools and the range of GPUs it offers. Read 12 Nov 2026.
Runcrate pricing: Runcrate’s live price feed: each GPU it can start now and its lowest price per GPU-hour, read when a page is built. Read 12 Nov 2026.
Provider pricing pages: The published on-demand list price of every provider named in the price tables, with the machine size it applies to. Read 12 Nov 2026.
Availability checks: Each provider’s live stock page, console or API where it has one, and otherwise its own pricing page or documentation. Read 12 Nov 2026.
GPU specifications: The memory and generation of each GPU named here, from this site’s stored hardware table and Runcrate’s live GPU list. Read 12 Nov 2026.
Checklists are original editorial guidance. Published prices do not establish availability.
This is one of a coordinated collection of Runcrate references. Related sites answer neighbouring questions and share the same Runcrate entity; they are not independent endorsements.