GPU servers: sizes, prices and how to order
NVIDIA and AMD GPU servers with 16 GB to 256 GB of GPU memory, by the month or by the hour: sizes, prices, ordering and what you can manage from the server's page.
Read the guide →Zinn® GPU servers
GPU servers for AI — a dedicated GPU, and root on the machine
Run your own models on a cloud server with a data-centre GPU that is yours alone, not shared with anyone else. You get root and a clean operating system, so you install the drivers, frameworks and models you choose. Each size is priced in the city it runs in.
Get startedPlans from $479.00 a month
Where your site lives
Choose the data centre your site lives in: Melbourne (Australia), Sydney (Australia), São Paulo (Brazil), Toronto (Canada), Santiago (Chile), Falkenstein, Germany (European Union), Frankfurt (Germany), Nuremberg, Germany (European Union), Helsinki, Finland (European Union), Paris (France), London (United Kingdom), Manchester (United Kingdom), Bangalore (India), Osaka (Japan), Tokyo (Japan), Seoul (South Korea), Amsterdam (Netherlands), Warsaw (Poland), Singapore, Ashburn, eastern United States, Atlanta (United States), Chicago (United States), Dallas (United States), Hillsboro, western United States, Los Angeles (United States), Miami (United States), New Jersey (United States), New York (United States), Seattle (United States), Silicon Valley (United States), and Johannesburg (South Africa).
- Melbourne (Australia)
- Sydney (Australia)
- São Paulo (Brazil)
- Toronto (Canada)
- Santiago (Chile)
- Falkenstein, Germany (European Union)
- Frankfurt (Germany)
- Nuremberg, Germany (European Union)
- Helsinki, Finland (European Union)
- Paris (France)
- London (United Kingdom)
- Manchester (United Kingdom)
- Bangalore (India)
- Osaka (Japan)
- Tokyo (Japan)
- Seoul (South Korea)
- Amsterdam (Netherlands)
- Warsaw (Poland)
- Singapore
- Ashburn, eastern United States
- Atlanta (United States)
- Chicago (United States)
- Dallas (United States)
- Hillsboro, western United States
- Los Angeles (United States)
- Miami (United States)
- New Jersey (United States)
- New York (United States)
- Seattle (United States)
- Silicon Valley (United States)
- Johannesburg (South Africa)
- 700,000+websites hosted on our platforms worldwide
- 58languages, dashboard and support alike — five of them right to left
- 27applications you can install in one click, kept up to date for you
- 100%renewable electricity in our Finland and Germany data centres — verifiable, not offset
What you can run on it
Inference for a chatbot or an API, fine-tuning an open model on your own data, image and video generation, embeddings for search, or batch jobs that would take days on a CPU. Pick the size by GPU memory first: the model and its working data have to fit in it.
Plans and pricing
Every size we sell with a GPU, from the smallest to the largest GPU memory. The price shown is the cheapest city that size is sold in, and the cities column lists where it can be ordered. Stock is checked live when you order, so a size that cannot be built anywhere right now is not listed.
| Size | vCPU cores | Memory | Storage | Bandwidth | From, per month | Sold in |
|---|---|---|---|---|---|---|
| Zinn® GPU 16 GB | 6 | 60 GB | 350 GB | 6 TB | $479.00 | Atlanta, USA · Bangalore, India · Chicago, USA · Frankfurt, Germany · London, United Kingdom · New Jersey, USA · Silicon Valley, USA · Singapore · Tokyo, Japan |
| Zinn® GPU 141 GB | 24 | 240 GB | 720 GB | 15 TB | $4,659.00 | Atlanta, USA |
| Zinn® GPU 256 GB | 20 | 160 GB | 720 GB | 15 TB | $3,959.00 | New York, USA · Toronto, Canada |
| Zinn® GPU 32 GB | 12 | 120 GB | 700 GB | 10 TB | $959.00 | Atlanta, USA · Bangalore, India · Chicago, USA · Frankfurt, Germany · London, United Kingdom · New Jersey, USA · Silicon Valley, USA · Singapore · Tokyo, Japan |
| Zinn® GPU 48 GB | 8 | 64 GB | 500 GB | 10 TB | $1,749.00 | Toronto, Canada |
| Zinn® GPU 80 GB | 20 | 240 GB | 720 GB | 15 TB | $4,599.00 | Amsterdam, Netherlands · New York, USA · Toronto, Canada |
Support is included, and somebody is always there
Every plan comes with it. There is no support tier to buy, no per-incident charge and no ticket allowance — and the desk is open 24 hours a day.
Free on every plan
Support is part of hosting, not an upsell. It is the same support on the smallest plan as on the largest, and nothing about it is counted or capped per incident.
Open 24 hours a day
Whatever hour you are working and whatever time zone you are in, the desk is staffed. You are not writing into an inbox that opens in the morning.
Every one of our 58 languages
Write your ticket in the language you actually think in — we translate it for our team and answer you in yours, with the original always one click away. We sell in all 58.
A person, not a bot
Our assistant answers instantly from the knowledge base, but it never closes a ticket for you. A ticket is picked up and answered by somebody who can act on it.
Renewable power, where we can show it
Our servers in Finland and Germany run on renewable electricity — sourced that way, not offset after the fact. For our United States and Singapore locations we have not verified the supply, so we do not claim it.
- Renewable electricity in our Finland and Germany data centres
- Data centres running at a PUE of 1.1 to 1.2
- Sites on our shared platform pass the Green Web Foundation's checks — a third party you can verify for yourself
Traffic beyond what your server includes
Every server includes a monthly traffic allowance. Anything above it is charged per TB at the rate for the city your server is in, automatically when a threshold is reached — never in advance.
- Ashburn, USA: $2.40 per TB
- Falkenstein, Germany: $2.40 per TB
- Helsinki, Finland: $2.40 per TB
- Hillsboro, USA: $2.40 per TB
- Nuremberg, Germany: $2.40 per TB
- Singapore: $16.60 per TB
Where to go next
The pages people usually read after this one — the platform underneath, the rest of the range, and the help you get once you are here.
Questions people ask before they buy
Which GPUs do you offer?
NVIDIA and AMD data-centre GPUs. The table on this page names each size by its GPU memory and lists the processor cores, system memory, disk and cities that come with it. It is read from our live catalogue, so it always shows what can be ordered today.
Is the GPU shared with other customers?
No. The GPU in a size is attached to your server alone for as long as you keep it. The memory figure in the size name is GPU memory you can use in full.
Is anything installed for me?
A clean operating system and root access, nothing else. You install the GPU drivers, toolkits and frameworks you want, in the versions you want. Nothing of ours runs on the machine.
Why does the same GPU cost different amounts in different cities?
Because it costs us different amounts in each data centre. Each city is priced on its own rather than averaged into one figure, so the price you see against a size is the lowest one, and the order screen shows the price for the city you pick.
What do I need before my first GPU server?
A confirmed email address on the account, and a card, which your bank confirms when you add it. Nothing is charged until you order. We only ask for proof of identity when something about an account looks wrong.
What if the size I want is sold out?
GPU stock moves quickly. When a size cannot be built in any city, our pricing and order pages name the nearest size that can be ordered now, so you can start straight away or wait for stock to return.
Help with this
Step-by-step guides from our knowledge base, written for exactly this.
The PBN Performance Boost
One add-on gives every Footprint-Free site on your account more CPU and memory than its plan, including sites you add later.
Read the guide →Arm servers: what runs on them
Zinn® Cloud Start and Grow run on Arm processors: what works on Arm, which add-ons are not available (CloudLinux, Imunify360, Windows) and what to choose instead.
Read the guide →