CLOUDTECHNOLOGIESNVIDIA GPUs
Eight NVIDIA GPUs per server on an HGX board, connected by NVLink for multi-GPU work.
AI servers on demand
GPU servers for AI training and inference.
Dell PowerEdge XE9780L servers with NVIDIA GPUs and direct liquid cooling, available on demand in our data center in the European Union.
XE9780LDell PowerEdge platform
NVIDIAHGX GPU platform
LiquidDirect liquid cooling
On demandCapacity on request
Overview
Built on Dell's liquid-cooled platform for AI.
Training, fine-tuning and serving large models takes GPUs. Our AI servers run on the Dell PowerEdge XE9780L, Dell's liquid-cooled platform for AI, with eight NVIDIA GPUs per server on an HGX board.
They are separate from our bare-metal and cloud servers, and offered on demand: tell us what you want to run and for how long, and we reserve the capacity for you.
Pricing depends on the capacity and term you need, so our team quotes every request.
Good for
- Model training
- Train models on multi-GPU servers built for sustained load.
- Fine-tuning
- Adapt foundation models to your own data.
- Inference at scale
- Serve large models to your users with low latency.
- Research
- Run experiments and benchmarks on dedicated GPUs.
What you get
What's included with AI servers.
Liquid-cooled Dell PowerEdge XE9780L servers with NVIDIA GPUs, available on demand for AI training and inference.
Liquid-cooled
Direct liquid cooling keeps GPUs and CPUs at full performance under sustained load.
Dell PowerEdge
Built on the Dell PowerEdge XE9780L, an enterprise platform designed for AI.
On demand
Reserve GPU capacity for the time you need it, from a single job to a longer project.
Hosted in the EU
Your data and models stay in our data center in Romania, inside the European Union.
Sized with you
Tell us about your models and data, and our team plans the right capacity.
Capacity
GPU capacity, on demand.
AI servers are reserved for the period you need. Tell us your workload and timeline, and our team will reply with availability and a quote.
- Platform
- Dell PowerEdge XE9780L
- GPUs
- NVIDIA HGX, 8 per server
- CPUs
- Intel® Xeon® 6
- Cooling
- Direct liquid cooling
- Availability
- On demand
- Pricing
- On request
How it works
Four steps from request to running.
Every server runs in our data center in Romania, inside the European Union.
Tell us your workload
Share what you want to run: models, data size and timeline.
We plan the capacity
Our team checks availability and sizes the GPU capacity you need.
We reserve your servers
Your AI servers are prepared in our data center in Romania.
Run your workload
Train, fine-tune or serve your models for as long as you need.
What are AI servers on demand?
Dedicated GPU servers that we reserve for you for the period you need. Tell us your workload and timeline, and our team arranges the capacity.
Which hardware do the AI servers use?
Dell PowerEdge XE9780L servers with direct liquid cooling and an NVIDIA HGX board with eight GPUs per server. Ask our team for the exact configuration.
How are AI servers different from bare metal?
AI servers are GPU servers for training and inference. Bare-metal servers are CPU servers with AMD EPYC™ processors for databases, heavy compute and CPU workloads. They are separate offerings.
How much do AI servers cost?
Pricing depends on the capacity and term you need. Request AI capacity and our team will reply with a quote.
More services
Our other services.
Bare metal
Single-tenant AMD EPYC™ servers with up to 128 threads and 512 GB of RAM for heavy compute and data.
Learn moreDedicated cloud
A scalable cloud with unshared vCPU power, ideal for production workloads.
Learn moreCloud hosting
Flexible cloud servers on high-end hardware for websites, shops and applications.
Learn more
Knowledge
Further reading.
Industry New
How AI Is Changing What Businesses Need From Their Hosting
AI is reshaping hosting requirements, from compute and memory to predictable performance, data location and energy efficiency.
Cooling
Cooling Solutions for Data Center Efficiency
From aisle containment to free cooling and direct-to-chip liquid systems, here is how data centers remove heat efficiently as rack densities rise.
AI New
Running AI Inference on CPUs: When It Makes Sense
Not every AI workload needs a GPU. Classical ML, embeddings, batch jobs and smaller quantized models often run well and cost-effectively on CPUs.
Get started
Tell us about your AI workload.
Client accounts are activated on request. Tell us what you need to run, and our team will set up your account and help you choose the right setup.