Your own GPU server, from 20 to 96 GB of card memory.
The whole machine answers to you, with full root access, a driver stack you choose and pin, and every hour of the card spent on your own work.
For inference, fine-tuning and rendering, by the hour or by the month.
- RTX 400020 GB VRAM
- NVIDIA A1024 GB VRAM
- NVIDIA L424 GB VRAM
- NVIDIA V100S32 GB VRAM
- 2 x NVIDIA V100S32 GB VRAM
- 4 x NVIDIA V100S32 GB VRAM
- NVIDIA L40S48 GB VRAM
- NVIDIA A10080 GB VRAM
- NVIDIA H10080 GB VRAM
- RTX PRO 600096 GB VRAM
From US$1.30 an hourFrom US$469.00 a month
- Traffic in and out of the machine is included
Pay by the hour, or by the month.
The same kind of card, two ways to run it. Pick the one that fits how you work.
By the Hour
For work that comes in bursts. Start it, run the job, and it stops when your credits do.
From US$1.30 an hour
See Hourly GPUsBy the Month
For work that runs around the clock, at one price for the whole month.
From US$469.00 a month
See Monthly GPUs
Pick the work. See what a month buys.
Choose the job and the size of your model. We show the smallest machine that holds it, its price from our catalogue, and what that buys in plain words.
Flare is your machine.
8 billion parameters at 16-bit works out at about 16 GB of weights.
4 GB to spareServingRun inference around the clock with the whole card to yourself, so the response times you measure on Monday are the response times your users get on Friday.Serving sits comfortably close to the size of the weights.
730 hours this month, all on your work
Tell us the model you are running and we will design the machine around it.
Talk to usChoose the machine that fits your model.
GPU by the Hour
Run a GPU only while you need it, billed by the hour from your credits.
- Billed by the hour from your credits.
- When your credits run out, the server is parked and charged for storage only.
- Top up your credits whenever you want more time on it.
Only the account owner can start a GPU server.
Powered off, a GPU server is still charged its full hourly rate. Park it to pay for its storage only; you see the storage charge before you park.
Which one do I need?
Three questions, one clear answer
Tell us what you're running. We'll recommend the server, the size and the location, at today's prices.
Ask Kora InsteadWhat are you running?
What people run on them.
Choose the card by its memory: your model, your batch or your scene has to fit on it.
Inference
Serve a model to your app or your users, with the whole card answering your requests and nobody else's.
Fine-Tuning
Train an open model on your own data, and keep every checkpoint on a machine only you can reach.
Rendering
Queue frames, video or embeddings and let the card run flat out until the job is done.
How our hourly price compares.
Each of our GPUs beside the same card where others sell it by the hour, with the memory and machine shown on both sides.
Ours
NVIDIA L4
24 GB VRAM, 0 vCPU, 0 GB memory
€1.16 an hour
Scaleway
L4-1-24G
24 GB VRAM, 8 vCPU, 48 GB memory
€0.79 an hour, on demand
As listed on 9 October 2026. Their Pricing Page
Ours
NVIDIA L40S
48 GB VRAM, 0 vCPU, 0 GB memory
US$2.30 an hour
DigitalOcean
GPU Droplet L40S x1
48 GB VRAM, 8 vCPU, 64 GB memory
US$1.57 an hour, on demand
As listed on 9 October 2026. Their Pricing Page
Ours
NVIDIA L40S
48 GB VRAM, 0 vCPU, 0 GB memory
US$2.30 an hour
Vultr
Cloud GPU L40S x1
48 GB VRAM, 16 vCPU, 180 GB memory
US$1.671 an hour, on demand
As listed in an archived copy of their page from 26 September 2026. Their Pricing Page
Ours
NVIDIA H100
80 GB VRAM, 0 vCPU, 0 GB memory
US$4.51 an hour
DigitalOcean
GPU Droplet HGX H100
80 GB VRAM, 20 vCPU, 240 GB memory
US$4.41 an hour, on demand
As listed on 9 October 2026. Their Pricing Page
Ours
NVIDIA A100
80 GB VRAM, 15 vCPU, 180 GB memory
US$4.50 an hour
Vultr
Cloud GPU PCIe A100 x1
80 GB VRAM, 12 vCPU, 120 GB memory
US$2.397 an hour, on demand
As listed in an archived copy of their page from 26 September 2026. Their Pricing Page
Their prices as listed, on demand, before tax. Ours are read from our catalogue today, in the same currency as theirs. Check both before you decide.
Your driver stack, pinned where you put it
Full root access and your own driver stack. Choose the driver and the toolkit, pin them together, and they stay exactly where you put them from one run to the next.
The whole machine, yours alone
Every hour the card works, it works on your job, and the machine answers only to you.
Who Looks After The Drivers
You do, on an unmanaged machine, and that is what most people here want. On a managed twin our engineers keep the operating system for you.
A clean Linux whenever you want one
Reinstall from the panel in a few clicks and the machine comes back with your own SSH keys already on it. Rescue mode boots at any hour, with your own keys.
Rescue mode, on your command
Boot into a rescue system from the panel at any hour, with your own keys, and take back control in minutes.
- On the card
- 20 GB
- In the machine
- 64 GB
- Operating System
- Linux
- SSH Keys
- 2 keys
A GPU server is a computer with a graphics card in it, and a graphics card is very good at doing the same small sum millions of times at once, which is exactly what training and running a model is made of. Renting one from us means the whole machine is yours: the card, the processor and the memory around them, all yours for as long as you keep it. You install whatever you like on it, and we keep it powered, connected and cooled.
Start with the memory on the card, because that is what sets how large a model you can load. Serving sits comfortably close to the size of the weights, and training rewards headroom above them, so give yourself room if you are training.
Choose the machine you want at the price on its card and we build it to order. Ask us where a tier stands on the day and we will tell you.
You do, on an unmanaged machine, and that is what most people here want. Pick the driver and the toolkit, pin them together, and they stay exactly where you put them. On a managed twin our engineers keep the operating system for you.
By the hour from your credits, or by the month at one price for the whole machine. A monthly machine renews at the price you took it at for as long as you stay on the plan.
The whole machine, yours alone. Your driver stack, pinned where you put it. A clean Linux whenever you want one: reinstall from the panel in a few clicks and the machine comes back with your own SSH keys already on it. Rescue mode, on your command. Traffic in and out of the machine is included.
An hourly GPU is billed from your account credits for each hour it runs. Add credits, start the server, and its hours are taken from your balance.
When your credits reach zero, the server stops by itself, so there is never a surprise bill. Top up your credits whenever you want more time on it.
Only the account owner can start a GPU server, and its hours are billed from the account's credits.
If you only power it off, yes: the full hourly rate. If you park it, only its storage is charged, and KPanel shows you that charge before you park. When your credits run out, the server is parked and charged for storage only.
Choose hourly for work that comes in bursts, like a training run or a batch of renders. Choose monthly for work that runs around the clock, like serving a model to your users.
You can order a GPU server in Frankfurt (Germany), Helsinki (Finland), London (United Kingdom) and France.
Tell us what you are running. Let's go.
- Traffic in and out of the machine is included