Data centres: liquid cooling, control valves, and Europe’s AI build out

Across Europe, investment is accelerating to build and upgrade facilities that can host AI and GPU dense compute. Vacancy has dropped to single digits and power is now the #1 constraint, pushing operators to refit existing rooms and develop new capacity in second tier markets.

Early rooms relied on industrial grade air conditioning and hot/cold aisle containment. That approach works well for ~2–20kW per rack—beyond that, room based cooling becomes inefficient and prone to hot spots.

AI servers have changed the envelope entirely. Depending on workload and platform, operators now plan for 20–120kW per rack—a density at which direct liquid cooling becomes not just preferable but essential.

From HVAC to liquid cooling (direct to chip)

For upgrades, a common retrofit path is adding chilled water piping and heat rejection via cooling towers or facility water loops. Heat is transferred from server cold plates into a secondary loop and then rejected outdoors via evaporation in the tower or through liquid to liquid exchangers. This path fits existing racks and dramatically increases heat removal per footprint—ideal for brownfield sites shifting HVAC → Direct to Chip (D2C) liquid cooling.

Cooling Distribution Units (CDUs): the heart of the system

A CDU is the control centre of D2C systems. It creates two isolated circuits:

• IT loop (to servers): pumps, manifolds, filters, and instrumentation deliver clean coolant to cold plates—maintaining tight setpoints above the dew point to avoid condensation.
• Facility loop (to plant): transfers IT heat to site chilled water or to a cooling tower via a liquid to liquid heat exchanger.

Modern CDUs provide redundant pumps (often N+1 or 2N), PID control of flow/pressure/temperature, quick disconnects, and water quality isolation—making them perfect for phased retrofits and mixed air/liquid halls.

Why valves matter in liquid cooled AI rooms

Valves sit on every critical branch—from the CDU to server manifolds and from the secondary loop to the facility plant. Their selection directly affects cooling reliability, uptime, and energy efficiency. In high density environments, operators are asking for:

• System integrity: leak tight, cleanable, with materials compatible with treated water / glycol mixes.
• Isolation valves: fast serviceability and safe zone isolation during maintenance.
• Precision control: smooth Cv characteristics for stable ΔT and flow.
• Chemical compatibility: elastomers and wetted metals suited to inhibitors and biocides.
• Ability to meet specs: documentation and options to satisfy site standards and OEM guidance.
• Tri clamp connections: frequently requested for sanitary, quick service interfaces.

We're aligned with suppliers actively supporting quick ship on many items and can scale deliveries for projects under tight timelines.

Typical CDU to rack valve architecture (at a glance)

• Facility loop (to tower/chiller): isolation ball valves, control valve for ΔT targeting, check valves at branch returns.
• IT loop (to cold plates): sanitary isolation valves (tri clamp), control valve for flow regulation per manifold, back pressure regulation to stabilise pump discharge, sight glass for visual checks.
• Instrumentation: temperature, pressure, and flow metering near the CDU and at manifold heads; keep IT loop supply above dew point to avoid condensation.

The wider market context we design against

• Rack densities are rising (AI/HPC) from ~15–30 kW to 60–120 kW in leading deployments; air cooling is generally inefficient above ~20 kW/rack.
• Europe’s capacity build out faces grid constraints, long equipment lead times, and sustainability pressures (including water usage considerations where towers are used).

How we help

If you're planning a retrofit or a new data centre:

• Valve specification & take off aligned to your CDU/OEM guidance.
• Material and elastomer selection for water glycol mixes and inhibitors.
• Sanitary connections (tri clamp) for quick maintenance in constrained rows.
• Quick ship options to de risk schedules.

Let’s map your rack densities, ΔT targets, and plant side constraints to a valve schedule that achieves reliability and efficiency—without slowing your build - contact the team today.