An AI GPU cloud: dedicated or shared GPU instances and an OpenAI-compatible AI API, hosted in France. The servers are immersed in a dielectric fluid that captures heat in direct contact with the components.
V100 32 GB
NVIDIA GPU
€0.18/h
starting from, excl. VAT
Cergy (France)
Hosted in France
Root SSH
key-based access
By the hour or by the month, no commitment, prepaid
Prepaid blocks of 5, 10, 25, 100 or 250 hours
Excluding VAT. VAT at next step
ConfigureVRAM budget overage: €0.02 excl. VAT per GB per hour
Deployment zone: Cergy AZ2
Prepaid blocks of 10, 25 or 250 hours
Excluding VAT. VAT at next step
ConfigureDeployment zone: Cergy AZ2
One prepaid month, no commitment
Excluding VAT. VAT at next step
ConfigureDeployment zone: Cergy AZ2
Prepaid blocks of 5, 10, 25, 50, 100 or 250 hours
Excluding VAT. VAT at next step
ConfigureDeployment zone: Cergy AZ2
The servers are immersed in a dielectric fluid, which does not conduct electricity. Heat is captured directly on each component, then carried away by a heat exchanger.
Immersion is the core business of VOLTANEUM by ITNET Technologies, part of the same group as Wayhost. VOLTANEUM designs and integrates the immersion tanks, the dielectric fluid, the heat exchangers (CDUs) and the container modules, operated in France and across Europe.
Dielectric fluid
it does not conduct electricity
Direct contact
heat is captured on each component
Heat exchanger
heat is carried out of the tank
The principle of immersion cooling, applied to your GPUs
The fluid captures heat right at the components, which helps GPUs keep a steady temperature during long computations.
A liquid carries heat far more efficiently than air: that is the very principle of liquid cooling.
Once immersed, components are no longer exposed to dust or ambient air.
Less dust, noise and vibration, and fewer hot spots around the servers.
| Criterion | Air cooling | Immersion cooling |
|---|---|---|
| Heat capture | Through the air, indirect | Direct, in contact with the components |
| What carries the heat | Air blown through the room | A dielectric fluid |
| Component exposure | Ambient air and dust | Immersed components |
Design targets and measurements, each labelled for what it is.
Energy efficiency
Target PUE of 1.03
for optimised high-density architectures
Density
200 kW and more
Up to 200 kW and more per tank depending on the integration (target)
Fans
No high-speed fans
inside the servers
Heat
Rejection or reuse
The heat, concentrated in the fluid, can be rejected or reused (water loop, heating) depending on the site.
Proprietary VOLTANEUM® dielectric fluid
Measurements on a tested batch.
Source: VOLTANEUM
OpenAI-compatible AI API
Keep your OpenAI API-compatible tools: OpenAI SDK, Open WebUI, LibreChat, n8n… Just set the Wayhost API address and your key. Each request is charged per token against your prepaid credit.
from openai import OpenAI
client = OpenAI(
base_url="API_BASE_URL",
api_key="YOUR_KEY",
)
r = client.chat.completions.create(
model="MODEL_ID",
messages=[{"role": "user", "content": "Hello!"}],
)
print(r.choices[0].message.content)Your computations are processed in France, and your AI API requests are never used to train models.
Need help choosing your GPU? Our team is here to answer.
Usually within minutes of payment: the IP address and the SSH command appear in your customer area.
The servers are immersed in a dielectric fluid, which does not conduct electricity. The fluid captures heat directly on the components, then a heat exchanger carries it out of the tank.
The instance is suspended and its data is kept for a limited time: top up hours from your customer area to restart it. After that period, it is deleted along with its data.
No. Your requests are not stored: only the model, the number of tokens and the cost are recorded for billing. Your application sends the conversation history with each call.
How does it work?
From the immersed server to the removed heat: immersion cooling in five steps.
Source: VOLTANEUM