Workstation AI Data Centre Solutions
Site acquisition, power and GPU procurement for data centre operators, AI start-ups and enterprises building AI capacity of their own. One partner from the first conversation to the first token.
What we do
Workstation supports data centre operators and AI infrastructure buyers end to end: site acquisition (powered land, available megawatts, colocation capacity), power and capacity planning, and GPU procurement for new builds and expansions. We handle the hardware refresh cycle too, buying and placing H100, H200 and similar 8-GPU systems.
We source to your spec, not from a fixed catalogue: new and refurbished NVIDIA and AMD GPU servers, networking, storage and the software stack, racked, burned in and handed over running.
When we share a live lot, we're representing our client.
Everything your data centre runs on
One supplier for the whole stack: accelerators from 32 GB to 1.4 TB of GPU memory, complete GPU servers, all-NVMe storage, 800 Gb/s networking, memory and the engineers to put it together. New or vetted refurbished, sourced to your spec.
VRAM decides which models you can run. Rule of thumb: about 1 GB per billion parameters at 8-bit, plus 20–30% for context and batching.
Per device or system. Prices ex VAT from the AI Shop; HBM systems are priced on quote.
H100, L40S and RTX PRO Blackwell Server Edition: passively cooled for rack airflow, new or vetted refurbished.
Browse GPUs →4U 8-GPU servers, HGX B200, DGX Station and DGX Spark, configured, burned in and delivered ready to run.
Browse servers →2U 24-bay all-NVMe data nodes, PCIe 5.0 U.2 drives up to 15.36 TB each, and 24 TB enterprise disks for cold data.
Browse storage →ConnectX-7 and ConnectX-8 adapters up to 800 Gb/s, BlueField-3 DPUs, Spectrum-4 Ethernet and Quantum-2 InfiniBand switches.
Browse networking →DDR5 ECC RDIMM kits up to 1 TB, Threadripper PRO and EPYC CPUs, and the GPUs to upgrade what you already own.
Browse components →On-site installation with a 72-hour burn-in, AI stack set-up and NVIDIA AI Enterprise licences.
See services →How we work
Site, power, GPU or configuration, quantity and timeline. The planner below gets you most of the way in two minutes.
We sign one before any details are exchanged on either side.
A quick, documented check so both sides know who they're dealing with.
We bring back options that fit your spec, budget and date, and work through next steps with you.
AI data centre planner · 7 quick questions
Answer seven questions. We build your profile as you go, then show the machines, the budget and the timeline that fit. No sign-up, no email.
为什么选择Workstation
Sites, power or GPUs. We start from what you need, not from what we happen to have.
We check supply before we bring it to you: provenance, condition and warranty.
NDA and KYC are standard steps, every time.
We rack it, burn it in for 72 hours and hand it over running your models.
Tell us what you need
Sites, power, GPUs or the whole thing: one conversation, one partner, one plan.
Details are shared under mutual NDA.