EstimatesHardware prices change daily, so prices here are estimates. Add what you need to your basket and request your final price: we reply within one working day.
Fit-checked machines · Engineers on the phone · Free trade accounts

AI workstations and GPUs for running models on your own hardware.

5 in UK stock · next-day
NVIDIA DGX Spark Founders Edition
128 GB unified · 273 GB/s
Price
£5,849 ex VAT
View product

For ML teams, labs and startups who'd rather own the box than rent the cloud. Every machine fit-checked here before it's listed, and the engineers who know these builds answer the phone.

ISO 27001

Trade accounts

Specified, quoted and procurement-ready.

A free Trade Account for every organisation. Pro adds an engineer and quote packs for £750 + VAT a year.

Free spec taster

Will it run? Size your build in 3 steps

What will you run?

Choose your account

Free

Trade Account

£0
  • Trade prices, shown ex VAT
  • Saved quotes with your PO number
  • Up to 3 users (we add them for you)
  • 5 AI engineer messages per chat
Open a free account
ProProcurement service

Pro

£750 + VAT / year
  • Engineer specs and sizing, in writing
  • Procurement-ready quote packs
  • 60-day price hold on your quotes while you get sign-off
  • Unlimited AI engineer chat and an engineer phone line (Mon–Fri 9–6)
Ask about Pro

Pro sign-up opens here soon. Ask us and we'll set it up.

Team / Institution

Several departments?

Fee by agreement
  • Several departments, or £50,000+ a year
  • A named contact
  • Users for every department
  • Everything in Pro
Talk to us

What your team's time on quotes costs

Based on your figures. Nothing is pre-filled.
0 h
Orders a year
£

Your figures

Set the hours per order, how many orders a year and your staff day rate to see what that time costs next to Pro.
Not sure which machine?

Start from what you'll run.

Pick the job and we've already picked the machine, each one fit-checked before it was listed.

Run a 70B model locally
ThinkStation PGX
128 GB unified. Quiet, desk-side, runs Llama 3.3 70B at Q4 with headroom.
Llama 3.3 70B Q4 · fit-checked
On quoteTalk to us
Office RAG box, 10 users
Mac Studio M5 Ultra 256 GB
Silent, 256 GB unified. Qwen 72B plus a vector index in memory, MLX or Ollama.
Qwen2.5 72B Q4 · fit-checked
£10,149ex VAT
Will it run?
Fine-tune on-prem
Supermicro 2× RTX 6000 Ada
96 GB GDDR6, CUDA. LoRA on 70B, full fine-tune to 13B. Assembled in Hemel.
2× RTX 6000 Ada · multi-GPU
£31,277.36ex VAT
Run DeepSeek-V3 at home
Mac Studio M5 Ultra 512 GB
Desk-side box that holds a 671B MoE at Q4 with room for context.
DeepSeek-V3 Q4 · fit-checked
On quoteTalk to us
Will it run?

Check a model against a machine before you buy it.

Fits in memory, headroom for context, and the speed to expect. Our bench numbers replace the estimate where we have them.

Quantisation

Machine

Fits
Llama 3.3 70B at Q4_K_M runs on the DGX Spark
16k context · 128 GB available · bandwidth estimate
Needs
58 GB
Headroom
+70 GB
Est. speed
3.8 tok/s · est.

Comfortable. 70 GB spare — room for 128k context, a second small model, or a RAG index in memory. Expect roughly 3.8 tok/s single-stream.

Machines that run this configuration

Not sure? Book a free 20-minute sizing call

Weights 42.8 GB at 4.85 bits/param · KV cache 5.0 GB at 16k (fp16, approximate) · 10 GB reserved for OS and runtime. Speed is single-stream decode, memory-bound; prompt processing and batching behave differently. Or WhatsApp an engineer on .

Trust

Real engineers, WhatsApp
· Mon–Fri 9–6 · tap to chat or call on WhatsApp · answered in Hemel Hempstead by people who benchmark these machines.
Returns you can plan around
Business customers can return unopened stock within 30 days; built-to-order systems only if faulty. Details in our Terms of Sale.
Next-day, UK stock
Order by 6pm. Stock counts on every product are live — if it says 4, there are 4 on the shelf.
Free to open, no contract
The Trade Account is free. Pro is annual: cancel from your account page and it runs to the end of the year you've paid for.