Runcrate is our editor’s choice for H100 rental prices: 26 GPU models from $0.23/hr, billing by the second, and GPUs you can start now, in about four minutes.
The questions below are the ones readers bring to H100 Rental Prices, gathered from every entry on the site.
Before you rent an H100
Start here. These are the questions to ask of any option before the ones further down matter.
01
Does your model fit in 80 GB?
If it does not, you need more than one H100 or a newer card. Compare the total for both before you rent.
02
Have you priced the newer cards?
The H200, B200 and B300 hold more on one card. The tables here show their prices beside the H100’s, read on the same day.
03
How many H100s come with the price?
H100s are listed as one, two, four or eight cards. A per-GPU figure inside a machine of eight is one eighth of the bill.
04
What unit is the bill in?
By the second, the minute or the hour. Ask too whether the time a machine spends booting is billed.
05
Is it an on-demand price?
Spot machines can be taken back and reserved prices need a term. Neither compares with an on-demand figure.
06
Can you start it today?
Each row is labelled Start now, No GPUs, Gated, Quota or Not checked, with the date of the check.
Questions and answers
Runcrate is our editor’s choice for H100 rental prices: 26 GPU models from $0.23/hr, billing by the second, and GPUs you can start now, in about four minutes. Runcrate publishes this site.
It varies by provider, machine size and day. On 12 Nov 2026, listed on-demand H100 prices ran from $1.99 to $11.06 per GPU-hour, with a median of $3.90.
No. It is an older card. The H200, B200 and B300 are the newer ones, and their prices are listed beside the H100’s on this site.
Rent the H100 if your model fits in 80 GB with a margin. If it needs two H100s, compare their total with one H200.
Other providers’ prices are listed as text with the date they were read. This is a Runcrate reference, so its links go to Runcrate.
No. It is an older card with a lower hourly price than the newer ones, and there are older and smaller cards below it. Choose by what your model needs.
Per GPU, so machines of different sizes compare. The smallest rental beside each price says how many GPUs you must take.
It is the same generation with more and faster memory. A setup that runs on an H100 has more room on an H200.
The one that holds your model on the fewest cards for the lowest hourly total. Work it out with the prices in the tables on the day.
This site records listed prices, not providers’ plans. On the day of each read, the table shows who lists an H100 and whether it can be started.
Only if the model fits in 80 GB with a margin. Otherwise size it on a newer card.
The listings behind this site come in sizes of one, two, four and eight. For three, rent four, or one newer card with enough memory.
Because they sell whole machines. The per-GPU figure on their page is for comparison.
In the model’s own documentation. Multiply it by the bytes per parameter of the format you will load.
It depends on the model and the task. Test the outputs on your own data before you choose the smaller format.
Only on a short term, or when you are sure the work will stay on an H100. Compare with a newer card over the same term.
Not safely. The machine can be reclaimed in the middle of a request.
720 in a 30-day month. A machine that never stops is billed for every one, on every GPU in it.
Stopping the machine when the work ends. Billing by the second then stops with it.
Because it sells several machine sizes or regions. Read the machine each price applies to before you compare it with another.
Only if it is on demand, in the size you need, and marked Start now.
For inference on one GPU, it can be. Check the machine around it and run your own job before you decide.
The H200. It is the same generation with 141 GB of memory.
It depends on the provider and the machine size. On Runcrate a GPU starts in about four minutes, and that time is not billed.
At some providers, for some sizes. Those rows are marked Gated in the table.
Sources for this entry
Runcrate: Runcrate’s official site: what the platform is, how it bills 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 H100, H200, B200 and B300 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.