On-demand
The server stays on for as long as you want at a fixed price. For training, serving a model and work that must not stop halfway.
Cloud GPU servers by the hour in Europe, the UK and the US
We provide NVIDIA GPU servers by the hour: H100, H200, B200, B300, RTX PRO 6000 and L40S in data centres in Finland, France, Spain, the UK and the US. The prices and capacity below refresh every few minutes.
What hourly rental means
You run a GPU server for as long as you need and pay only for the hours it is on. The server sits in a data centre in the country you choose, and your data stays there. Prices are shown in dollars and in Toman; Toman is worked out at the live free-market dollar rate and rounded up, and the final amount is fixed in the quote.
Three kinds of server
Spot
Uses the data centre's spare capacity and costs several times less, but can be taken back at short notice, and its price moves with supply and demand. For tests and jobs that can restart.
Eight-GPU servers and clusters
A whole server with eight GPUs and fast links between them for large models, and larger clusters or multi-month reservations at an agreed price.
Live GPU server prices and capacity
Updated 23 Oct, 09:04, Tehran time· Free-market dollar: 271,905 Toman
Each GPU-hour price is rounded up to the next whole dollar and its Toman figure to a round number. The final amount is the one in your quote.
| Country | Available now | On-demandper GPU-hour | Spotper GPU-hour | Order |
|---|---|---|---|---|
| NVIDIA B300270 GB memory · 24 vCPUs and 346 GB RAM per GPU | ||||
| France |
| $133,540,000 Toman | $133,540,000 Toman | FormWhatsApp |
| United Kingdom |
| $133,540,000 Toman | $133,540,000 Toman | FormWhatsApp |
| US, Minnesota |
| $133,540,000 Toman | $133,540,000 Toman | FormWhatsApp |
| NVIDIA B200180 GB memory · 20 vCPUs and 224 GB RAM per GPU | ||||
| US, Kansas City |
| $123,270,000 Toman | $41,090,000 Toman | FormWhatsApp |
| NVIDIA H200141 GB memory · 16 vCPUs and 200 GB RAM per GPU | ||||
| Finland |
| $82,180,000 Toman | $41,090,000 Toman | FormWhatsApp |
| France |
| $82,180,000 Toman | $82,180,000 Toman | FormWhatsApp |
| US, Kansas City |
| $82,180,000 Toman | $2550,000 Toman | FormWhatsApp |
| NVIDIA H10080 GB memory · 16 vCPUs and 200 GB RAM per GPU | ||||
| Finland |
| $61,640,000 Toman | $2550,000 Toman | FormWhatsApp |
| NVIDIA RTX PRO 600096 GB memory · 24 vCPUs and 218 GB RAM per GPU | ||||
| Spain |
| $3820,000 Toman | $2550,000 Toman | FormWhatsApp |
| United Kingdom |
| $3820,000 Toman | $2550,000 Toman | FormWhatsApp |
| US, Kansas City |
| $3820,000 Toman | $2550,000 Toman | FormWhatsApp |
| NVIDIA L40S48 GB memory · 8-40 vCPUs and 32-160 GB RAM per GPU | ||||
| Finland |
| from $3820,000 Toman | Contact us | FormWhatsApp |
Spot price changes over the last 30 days
Spot prices move with supply and demand; the chart shows our price for each country, hour by hour.
- France
- United Kingdom
- US, Minnesota
Dollars per GPU-hour, rounded up to the next whole dollar
On-demand price changes
We record the on-demand prices and the dollar rate every day; this history starts on 22 October 2026.
- NVIDIA B300
Dollars per GPU-hour
From order to a running server
You send the order by form or WhatsApp and we hold the capacity for you. The quote says when access will be ready.
- 01
Choose
Pick the GPU, the number of cards, the country and the type (on-demand or spot) in the live table and press order.
- 02
Capacity and quote
We check capacity at that moment and send a quote at the day's rate.
- 03
Payment
You pay the quote for the hours you need; for longer use we extend it the same way.
- 04
Access
We hand over the server with its operating system and GPU drivers ready, and send SSH access.
- 05
Support
If you want, we also set up the model and the assistant on it and train your team.
How to get in touch
Fill in the form or message us on WhatsApp. The clearer these are, the more exact the quote:
- What the model has to do (answer from documents, code, summarise, process images or speech)
- How many people use it at the same time, and what response time is acceptable
- The model or hardware family you have in mind, if you already know
- Where it goes: your server room, a data centre, or hardware you already own
- How sensitive the data is, and whether it must run with no internet at all
We reply within one business day.
Zharf
Try the ZharfAI assistant right now
Ask about Persian assistants, document automation, or how to start a pilot.
Frequently asked questions about renting GPU servers
How are Toman prices worked out?
We round the dollar price per GPU-hour up to the next whole dollar, convert it at the live free-market dollar rate, and round the Toman figure up. Because the rate moves, the final amount is the one in your quote.
What is the difference between on-demand and spot?
An on-demand server stays on as long as you want. A spot server uses spare capacity, costs less and changes price every few minutes, but can be stopped if the capacity is needed. Choose on-demand for work that must not stop.
What does "available now" mean in the table?
It is the number of servers we can start for you in that country right now, and the label next to it shows the chance of getting one. The figures refresh every few minutes; for more machines or a multi-month reservation, contact us.
Where are the servers, and where does data stay?
In data centres in Finland, France, Spain, the UK and the US. Data stays on the server in the country you chose, and we do not move it anywhere else.
Which GPU do I need for which model?
It depends on the model's size. The "which model on which hardware" table on the AI infrastructure page is a quick guide; DeepSeek-V4-Flash, for example, fits on one eight-GPU H100 server or four H200s. If you are unsure, we suggest one in the first conversation.
Does it work without international internet?
Yes, if the model, the serving software, the assistant, user sign-in and monitoring all sit inside your network. Anything that depends on a foreign service stops during a cut, so before handover we test the system with international internet deliberately cut off.
Pick the GPU, leave the rest to us
Order the row you want from the table above, or tell us what you want to run and we will suggest the right server and country.