DataCentersX > Stack > Cooling and Thermal Management > Cooling Tower and Heat Rejection

DC Cooling Tower and Heat Rejection

Heat rejection is the terminal stage of the thermal chain. Every watt of electrical power delivered to a data center has to leave the building as heat, and the mechanical plant is where that transfer finally happens: the facility water loop surrenders its heat to atmosphere through a cooling tower, dry cooler, or hybrid system, or in a small but growing minority of sites, to a district heating network or industrial reuse partner. This page covers the engineering of the terminal rejection stage, regardless of what cooling modality (air, liquid, DTC, immersion) feeds the facility loop upstream.

The central engineering decision in rejection design is the balance between water consumption and electrical energy. Evaporative cooling is thermodynamically efficient but consumes water continuously. Dry cooling consumes no water but requires higher approach temperatures and more compressor work. Hybrid systems combine both and dominate new hyperscale design at sites where either extreme is unacceptable.

Rejection technology classes

Technology	Heat Transfer Mechanism	Approach Temperature	Water Use
Open-circuit cooling tower	Evaporation of tower water into passing airstream	Low; 3 to 5 degrees C above ambient wet-bulb	High; evaporation plus drift plus blowdown
Closed-circuit fluid cooler	Evaporation on tube exterior; process fluid in sealed coil	Low; 4 to 7 degrees C above ambient wet-bulb	Moderate; evaporation on tubes only
Dry cooler	Finned-tube heat exchanger; air-side convection only	High; 8 to 12 degrees C above ambient dry-bulb	Zero
Hybrid (adiabatic) cooler	Dry at low ambient; water pre-cooling of inlet air at high ambient	Variable; approaches wet-bulb in adiabatic mode	Seasonal; minimal in cool months, moderate in peak heat
Mechanical chiller	Vapor-compression refrigeration; compressor work added	Arbitrary; set by chiller control	Depends on condenser type (tower or dry)

Open cooling towers

An open-circuit cooling tower rejects heat through evaporation. Warm water from the facility loop is distributed over fill media at the top of the tower; air drawn upward by a fan passes through the falling water; a fraction of the water evaporates into the air, absorbing latent heat and cooling the remainder. Cool water collects in the basin and returns to the loop. The tower water is in direct contact with atmosphere, which is why the facility loop it serves is the facility water loop rather than the technology cooling loop — TCS-grade chemistry cannot survive open-tower exposure.

Open towers achieve the lowest approach temperatures of any rejection technology because evaporation exploits the wet-bulb temperature, which is typically several degrees below dry-bulb in any humidity short of saturated air. This is the thermodynamic reason evaporative cooling has historically dominated large mechanical plants: for a given heat load, a tower needs less fan power and less heat exchanger surface than a dry cooler delivering the same approach.

The water cost is continuous and non-trivial. Evaporation removes pure water and leaves solutes behind, concentrating the tower basin. Blowdown (periodic partial drain and refill with makeup water) keeps solute concentration below scaling or corrosion thresholds. Drift (droplets carried out by airflow) adds a smaller but real loss. At hyperscale, a single campus can consume millions of liters per day through these mechanisms.

Dry coolers

A dry cooler is an outdoor finned-tube heat exchanger with fans. The process fluid passes through the tubes; ambient air passes over the fins; heat transfers by sensible convection only, with no phase change and no water consumption. The approach temperature is fundamentally higher than an evaporative tower because dry cooling exploits dry-bulb rather than wet-bulb temperature, and the sensible heat capacity of air is much lower than the latent heat capacity of evaporating water.

The consequences of the higher approach ripple through the rest of the cooling plant. Chilled water supply temperature has to be higher, or the mechanical chillers have to work harder to maintain the target supply. In hot ambient conditions, a dry-only plant may have to lift supply temperatures into ranges that force reduced IT load or violate cooling contracts. This is why dry cooling is common in cool climates (northern Europe, the Pacific Northwest) and uncommon in hot ones, at least without supplementary evaporative capacity.

The offsetting advantage is water independence. Dry-cooled facilities can be sited without reference to water availability, without permitting for withdrawal or discharge, and without exposure to drought restrictions. For operators in water-stressed regions (the US Southwest, parts of Spain, Australia) or for facilities under jurisdictional pressure on water consumption, dry cooling increasingly wins even at the cost of higher PUE.

Hybrid and adiabatic systems

Hybrid systems combine dry and evaporative operation in a single unit. In cool ambient conditions the system runs as a dry cooler, consuming no water. As dry-bulb temperature rises past a setpoint, pre-cooling sprays wet the inlet air or the heat exchanger tubes, dropping the effective approach toward wet-bulb. The transition is often continuous: more water use as ambient gets hotter, zero water use in shoulder seasons and winter.

The operational profile matches what most sites actually need. Cooling load peaks in hot weather when the grid is stressed and mechanical chillers are expensive to run; hybrid systems buy lower approach temperatures exactly during those hours, avoiding the PUE penalty of pure dry cooling at peak. At the same time, winter and shoulder-season operation consumes little or no water, which reduces annual water usage dramatically relative to an open tower.

Hybrid cooling is emerging as the default rejection technology for new hyperscale builds in any region where water availability is constrained or regulated. It is more expensive than either extreme in capital cost, but the combination of low annual water consumption and acceptable peak-season PUE makes the TCO math work at sites where pure evaporative is no longer permitted.

Mechanical chillers

Chillers are vapor-compression refrigeration systems that add compressor work to move heat from a cooler source (chilled water loop) to a warmer sink (condenser loop). A cooling plant uses chillers when the ambient sink is too warm to cool the chilled-water supply to target without them. The chiller does not eliminate the need for rejection; it shifts heat from the chilled-water loop to a condenser loop that still has to be rejected by a tower, fluid cooler, or dry cooler.

The design variable that most affects chiller energy is how many hours per year the plant can bypass the chillers entirely and run the condenser loop directly into the chilled-water supply. This is the engineering meaning of free cooling or waterside economization: when ambient is cool enough that the tower or dry cooler alone can deliver chilled-water supply at target, compressors idle and the plant runs on fan and pump power only.

At modern direct-to-chip densities with supply temperatures in the 30 to 45 degree C range, chillers are increasingly unnecessary for large parts of the year even in warm climates. This is the mechanism by which DTC-native facilities reach PUE values below 1.1: the chiller runs few hours, the rejection side runs ambient, and the compressor energy that dominated traditional hyperscale plants collapses.

Waste heat reuse

Every rejection technology above treats data center heat as waste to be dumped. An alternative treats it as a product to be delivered to an adjacent consumer. Waste heat reuse diverts some or all of the facility loop's heat to an external network, most commonly a district heating system serving nearby buildings.

The economics and physics are constrained. District heating networks in most jurisdictions run supply temperatures in the 60 to 90 degree C range, and data center facility water loops run much cooler (35 to 50 degrees C at best). Delivering useful heat requires either a high-temperature loop (which penalizes cooling performance) or heat pumps to lift the data center return to district supply temperature (which costs electrical energy). Both options are viable in the right geographic and regulatory context and both have been deployed at scale in Scandinavia, the Netherlands, and parts of Germany and Switzerland.

Outside regions with established district heating infrastructure, waste heat reuse remains a small share of total data center heat disposal. Industrial reuse (greenhouses, aquaculture, process heating for adjacent tenants) is a niche pattern that works when a large heat consumer happens to locate nearby. Regulatory and economic pressure is building in the EU (energy efficiency directives, heat network mandates) to make reuse a default rather than an exception for new facilities above a capacity threshold.

Site-level selection

Rejection technology selection is a site-level decision driven by three variables: local climate (dry-bulb and wet-bulb profiles across the year), water availability and cost (physical, regulatory, and reputational), and IT load temperature profile (DTC-native loads tolerate higher supply temperatures and enable more free cooling than air-cooled halls). The combination of these variables determines whether a site runs open tower, hybrid, dry cooler, or some combination, and whether it can target sub-1.1 PUE or has to accept the penalty of compressor-heavy operation.

The trend in new hyperscale design is toward higher facility loop temperatures (enabled by DTC), hybrid or dry rejection (enabled by higher supply temperatures), and waste heat reuse where geography permits. The combined effect collapses both compressor energy and water consumption relative to legacy air-cooled plants running at low chilled-water temperatures.

Related coverage