GPU & AI COMPUTE
Dedicated GPU & AI Servers, hosted in the UK.
CUDA ready accelerated compute on single tenant Dell hardware, plus a desktop class Grace Blackwell system for local AI development. Predictable monthly pricing, no shared neighbours, no cloud GPU surge bills.
WHAT THEY ARE BUILT FOR
Right sized accelerated, honestly described.
These are dedicated NVIDIA Telsa V100S servers - proven, cost effective Volta-generation accelerators with 32 GB of memory per GPU. They earn their keep on the workloads below, where consistent throughput and a fixed monthly cost matter more than chasing the newest silicon.
Model inference
Serve trained models in production with dedicated, always on GPU capacity.
Fine tuning
Adapt and fine-tune smaller and mid sized models without renting cloud GPUs by the hour.
CUDA dev & test
A stable, dedicated environment for CUDA development, CI and reproducible benchmarking.
Render & VDI
GPU accelerated rendering, transcoding and virtual desktop workloads.
Where they fit, and where they don't
V100S accelerators are excellent value for inference, fine tuning and accelerated compute. They are not intended for training large frontier scale language models from scratch - if that is your goal, talk to us about requirements before you commit, and we will tell you straight whether this is the right tool or if we can offer you the right tool for the job.
DELL GPU SERVERS
Single tenant V100S servers.
Built on Dell PowerEdge hardware with dual Intel Xeon Gold 5318Y processors (24 cores / 48 threads). Every server includes 2x 480 GB SSDs for the operating system, with high speed NVMe class SSD storage for your data and scratch.
Choose your GPU count, memory and storage.
Dual GPU
2 × NVIDIA V100S
64 GB total GPU memory · 2× Xeon Gold 5318Y · 24C / 48T · 30 TB bandwidth · UK data centre
| Configuration | Monthly |
|---|---|
| 256 GB RAM · 2× 3.84 TB SSD data | £599 /mo+ VAT |
| 256 GB RAM · 4× 3.84 TB SSD data | £628 /mo+ VAT |
| 512 GB RAM · 2× 3.84 TB SSD data | £688 /mo+ VAT |
| 512 GB RAM · 4× 3.84 TB SSD data | £717 /mo+ VAT |
Quad GPU
4 × NVIDIA V100S
128 GB total GPU memory · 2× Xeon Gold 5318Y · 24C / 48T · 30 TB bandwidth · UK data centre
| Configuration | Monthly |
|---|---|
| 256 GB RAM · 2× 3.84 TB SSD data | £899 /mo+ VAT |
| 256 GB RAM · 4× 3.84 TB SSD data | £928 /mo+ VAT |
| 512 GB RAM · 2× 3.84 TB SSD data | £988 /mo+ VAT |
| 512 GB RAM · 4× 3.84 TB SSD data | £1,017 /mo+ VAT |
AI DEVELOPMENT SYSTEM
A Grace Blackwell box for local AI work.
Different job, different machine. The NVIDIA DGX Spark is a self contained Arm based AI computer - not a rack server - aimed at one developer or a small team who want to prototype, fine tune and run reasoning models locally before scaling out.
NVIDIA DGX Spark
Platform: NVIDIA GB10 Grace Blackwell Superchip · DGX OS (Arm)
Coherent unified memory means the CPU and GPU share one 128 GB pool, so you can hold large models in memory and iterate fast. Link two units together for models up to ~405 billion parameters.
- CPU · 20-core Arm (10× Cortex-X925 + 10× Cortex-A725)
- GPU · NVIDIA Blackwell, 5th-gen Tensor Cores
- Memory · 128 GB LPDDR5x unified
- Storage · 4 TB NVMe
- Networking · 10 GbE + NVIDIA ConnectX
- Bandwidth · 30 TB · UK
1 PFLOP · AI performance (FP4) · up to 200B params · £199 /mo+ VAT
INCLUDED WITH EVERY SERVER
The basics, sorted.
TENANCY
Single tenant
BANDWIDTH
30 TB included
LOCATION
UK data centre
HARDWARE
Dedicated, not shared
Need more bandwidth, DDoS protection or a managed layer on top? We can add those - ask when you enquire.
Questions
GPU and AI hosting, answered.
Inference, fine tuning of small and mid sized models, CUDA development and testing, and GPU accelerated rendering, transcoding and virtual desktops. Each server gives you dedicated NVIDIA V100S GPUs with 32 GB of memory each, so throughput is consistent and the monthly cost is fixed.
For training large, frontier scale models from scratch, V100S hardware is not the right choice - newer accelerators are better suited to that. V100S servers are well matched to inference and to fine tuning smaller and mid sized models. Tell us your workload and we will say honestly whether this is a good fit.
The DGX Spark is a self contained Arm based AI computer built on NVIDIA's GB10 Grace Blackwell platform, aimed at a single developer or small team prototyping and fine tuning models locally. The V100S servers are dual Xeon rack servers for production inference and team workloads. Different machines for different jobs.
No. Every GPU server is single tenant - dedicated hardware, not a shared or virtualised slice. The resources are yours alone.
In our flag ship UK tier 3 data centre. Every server includes 30 TB of bandwidth as standard, with more available on request.
A fixed monthly price plus VAT, with no hourly metering or surge charges. Add-ons such as extra bandwidth, DDoS protection or a managed layer can be included - just ask when you enquire.
Tell us the workload, we'll size the server.
Send us what you're running and we'll recommend the right configuration or tell you honestly if it isn't a fit.