Proposed reference configurations
The standard progression starts with one office workstation, then moves into increasingly capable 4U rack systems. A dual-GPU tower remains available only as a specialist alternative when a rack cannot be used.
Standard progression
Price and physical infrastructure now move in one direction. The range does not return to a tower after the first rack server.
/01 GPU route
Standard rangeOne 32GB GPU for private document assistants, coding, transcription, image work and a measured office proof of value.
1 × NVIDIA GeForce RTX 5090 32GB
/02 GPU route
Standard rangeTwo independent 24GB GPU workers for private document assistants, coding, rendering and measured proof-of-value projects.
2 × refurbished NVIDIA RTX A5000 24GB
/04 GPU route
Standard rangeA dense multi-GPU platform for concurrent private assistants, mixed AI services, render queues and supported model sharding.
4 × proposed passive RTX 5090-class 32GB GPUs
/02 GPU route
Standard rangeProfessional server GPUs, modern PCIe 5.0 platform and the memory headroom needed for larger private inference and high concurrency.
2 × RTX PRO 6000 Blackwell Server Edition 96GB
/04 GPU route
Standard rangeA flagship four-GPU server for AI consultancies, engineering teams and organisations running several large private services.
4 × RTX PRO 6000 Blackwell Server Edition 96GB
Proposed indicative price, excluding VAT. Final price, specification, availability, delivery, warranty and model fit require a current supplier quote and written order confirmation.
Specialist workstation · outside the ladder
Company 64 is retained for a buyer who needs two current 32GB GPU workers but genuinely cannot install a rack. Its £37,000 indicative price reflects expensive workstation components and the 60% margin model; it is not presented as more physically capable than the rack platform below it.
The standard range will not invent a replacement product or price. A alternative-supplier rack family can be added here only after supplier, configuration, warranty, image and fully loaded cost evidence are recorded.
/02 GPU route
Specialist alternativeA separately evaluated Threadripper workstation for concurrent private AI, development, rendering and batch queues, with 64GB aggregate GPU memory.
2 × NVIDIA GeForce RTX 5090 Founders Edition 32GB
Standard range side-by-side
| System | Form | GPUs | Per-GPU VRAM | Aggregate VRAM | System RAM | Indicative ex VAT |
|---|---|---|---|---|---|---|
| Team 32 | Workstation | 1 | 32GB | 32GB* | 64GB | £14,500 |
| Core 48 | Rack server | 2 | 24GB | 48GB* | 128GB | £27,500 |
| Business 128 | Rack server | 4 | 32GB | 128GB* | 256GB | £70,500 |
| Sovereign 192 | Rack server | 2 | 96GB | 192GB* | 256GB | £97,000 |
| Scale 384 | Rack server | 4 | 96GB | 384GB* | 512GB | £172,000 |
A workload must fit on one GPU, use supported parallelisation or be split into independent workers. Context, cache, batch size and concurrency also consume memory.
Company 64 is deliberately excluded from this progression. See the specialist workstation above when two GPUs are required without a rack.
Included baseline
The reference price covers the agreed physical build, initial software profile and documented remote onboarding boundary. It is not a promise of open-ended implementation or a managed service.
Do not choose from a card
The right outcome may be a smaller machine, cloud service, colocation, hybrid route - or no purchase.