Compare & price
Home › Questions › High-density and GPU colocation: what are the constraints?

High-density and GPU colocation: what are the constraints?

Direct answer

A modern GPU server draws 3 to 10 kVA on its own: the classic rack (2–6 kVA) can’t keep up. You need high-density rooms (10–40 kVA/rack), often liquid cooling, and pricing negotiated per kW.

AI has blown up densities: a 4–8 GPU chassis draws 3 to 10 kVA, a full compute rack easily exceeds 30–40 kVA — ten times the classic enterprise rack. Consequences:

So state kVA/rack and cooling mode in your very first inquiry — it is the first filter, before price.

Current benchmarks from our catalog (updated 2026-11-01)
Cross-connect90–170 €/mo
IP transit 1 Gbps150–400 €/mo
Quarter rack250–500 €/mo
Power150–260 €/kVA/mo
IPv4 address2–3 €/IP/mo

Price my architecture →   Compare data centers

Related questions

← All questions

Advertisement