Investment | Price benchmarks | 04
Server rack
At rack level, the good decisions stop being about components and start being about density: how much heat you can remove, how much power you can bring in, and which network topology keeps your nodes busy. This is where a system either scales or stalls.
Typical configuration
CPU / virtualisation; AI rack (2–4 nodes + fabric)
Investment range
€250–900k | AI: €1.0–3M
What this tier involves
- Typical role
- Team HPC, private cloud, compact AI pod
- Density
- From ~10 kW/rack classic up to 40–120 kW for AI
- Facility
- Requires aisle containment, correctly sized PDUs & often a CDU
- Fee
- Basic €1.5–8k | Platinum €15–120k | Supreme €12–35k/year
Indicative figures, August 2026, excluding VAT. High volatility on GPUs and memory — firm quotes are valid for 30–60 days. The final price is set at scoping.
What matters at this scale
At rack level the power, heat and weight budgets are set, and the heat-removal path and network topology are chosen.
- The rack budget
- Power, heat and weight are budgeted per rack; permissible floor loading is verified before installation
- The thermal path
- Aisle containment, heat-exchanger doors or cold plates — selected according to the designed density
- Distribution & protection
- PDUs sized per phase and verified breaker selectivity, so faults are isolated at node level
- Starter fabric
- The network topology is chosen with port and bandwidth reserve for extension to a full row
- Telemetry
- Power, inlet temperatures and events collected per rack, with versioned alarm thresholds