❄️ Chilled Water vs. DX (Direct Expansion) Systems: Which HVAC Cooling Approach to Choose

The Fundamental Difference

Every mechanical cooling system ultimately does the same thing — moves heat from inside a building to outside it using a vapor-compression refrigeration cycle. The question is where that refrigeration cycle happens and what medium carries the cooling effect to the occupied space. A DX (direct expansion) system puts the evaporator coil directly in the airstream being conditioned — refrigerant expands and evaporates right where the air is cooled, whether that's a rooftop unit, a split system condensing unit paired with an indoor fan coil, or a packaged terminal unit. A chilled water system separates the two steps: a central chiller runs the refrigeration cycle to produce chilled water (typically 42–45°F supply), and that water is pumped through a piping distribution network to remote air handling units (AHUs) or fan coil units, where it absorbs heat from the airstream via a cooling coil before returning to the chiller.

This one architectural choice — refrigerant-to-air directly, versus refrigerant-to-water-to-air — cascades into nearly every other difference between the two approaches: first cost, redundancy, efficiency curve, maintenance access, refrigerant safety code compliance, and how well the system scales.

First Cost and Scale Thresholds

DX systems win on first cost at small-to-mid scale. A rooftop DX unit or a handful of split systems can be installed with minimal engineering, no central plant room, no chilled water piping distribution, and no pumps — for a building under roughly 50,000–75,000 sq ft, or any building without a strong central plant business case, DX is usually the lower first-cost path and often the only economically justifiable one.

Chilled water systems carry a substantial fixed cost premium — a central chiller plant, primary/secondary (or variable primary) pumping, cooling towers or air-cooled condensers, water treatment, and a piping distribution loop — that has to be justified by scale. The crossover point where chilled water starts to beat DX on total installed cost is typically cited around 100–150 tons of total building cooling load, though this varies significantly with climate, number of buildings served (a chilled water plant can serve multiple buildings from one central plant — a campus or hospital), and local labor/equipment costs. Below that threshold, the fixed plant cost isn't amortized over enough capacity to compete with simply installing more DX units.

Energy Efficiency and Part-Load Performance

At full load, well-selected DX and chilled water systems can have comparable rated efficiency — DX equipment efficiency is commonly expressed as EER or SEER2 (post-2023 DOE test procedure), while chiller plant efficiency is expressed in kW/ton (a good water-cooled centrifugal chiller can achieve 0.5–0.6 kW/ton at full load; air-cooled chillers typically run 0.9–1.2 kW/ton). The real efficiency gap opens up at part load, which is where commercial buildings spend the vast majority of their operating hours (ASHRAE load profiles show most buildings operate below 50% load more than 70% of the year).

• DX part-load penalty — most DX equipment modulates capacity through compressor staging or, at best, a single variable-speed compressor per unit; each independent DX unit has its own fixed losses (fan power, cycling losses, minimum compressor turndown) that don't scale down gracefully, and IPLV (Integrated Part Load Value) ratings for packaged DX equipment are frequently overstated relative to real-world performance.
• Chilled water part-load advantage — a central plant with multiple chillers can stage entire chillers on and off to match load, each one running near its own best efficiency point, and variable-primary or variable-speed chiller plants can achieve NPLV (Non-standard Part Load Value) efficiencies well below 0.4 kW/ton at partial load — something no single packaged DX unit can approach. This is the single biggest reason large buildings default to chilled water: the plant-level optimization opportunity doesn't exist in a fleet of independent DX boxes.

Pump and fan energy partially offsets this advantage — chilled water systems add continuous pumping energy (typically 15–25 W/ton for a well-designed variable-flow system) that a DX system doesn't carry, so the net efficiency comparison has to include the full system, not just the refrigeration cycle in isolation.

Redundancy and Reliability

Chilled water plants offer a structural redundancy advantage: with N+1 chiller sparing (e.g., three 500-ton chillers sized so any two can carry the full 1,000-ton design load), a single chiller failure or scheduled maintenance event doesn't take down cooling to any part of the building — the remaining chillers simply carry more load through the shared distribution loop. DX systems, by contrast, are inherently distributed and self-redundant in a different way: a failed rooftop unit takes down only the zone it serves, while every other zone continues operating normally. Neither architecture is unambiguously "more reliable" — chilled water offers centralized N+1 sparing for the whole building at the plant level, while DX offers built-in fault isolation at the zone level. Mission-critical facilities (hospitals, data centers) commonly use chilled water specifically because centralized N+1 (or 2N) plant redundancy is easier to engineer, monitor, and maintain than coordinating redundancy across dozens of independent DX units.

Refrigerant Charge and Code Implications

This is an increasingly important differentiator as refrigerant regulations tighten. A DX system's entire refrigerant charge sits inside occupied-space equipment (or close to it) — every fan coil, every rooftop unit, every zone has refrigerant piping running to or through occupied areas. A chilled water system confines all refrigerant to the central plant room; the distribution network running throughout the building carries only water, which has no flammability, toxicity, or GWP (global warming potential) concerns and no ASHRAE 15 refrigerant safety group restrictions in occupied spaces.

This matters directly for code compliance as the industry transitions to lower-GWP refrigerants. Many next-generation refrigerants (R-454B, R-32) are classified A2L — mildly flammable — under ASHRAE 34, and ASHRAE 15-2022 imposes specific refrigerant concentration limits and charge-size restrictions per occupied space for A2L systems. Large-capacity DX systems using A2L refrigerants can hit these limits, especially in dense occupancies, in ways that push designers toward either splitting equipment into more, smaller units or toward reconsidering chilled water specifically to keep refrigerant charge confined to the mechanical room. This is a live, evolving design consideration, not a settled historical fact — expect refrigerant-charge-driven system selection to become more common through the late 2020s as the industry works through the A2L transition.

Maintenance, Controls, and Operational Complexity

DX systems, especially packaged rooftop units, are mechanically simpler and more familiar to a broader base of HVAC service technicians — a technician who services one manufacturer's packaged units can generally service any building using that same standard equipment, and failures tend to be self-contained and diagnosable without specialized plant operations knowledge.

Chilled water plants require a materially higher level of operational sophistication: chiller sequencing logic, condenser/cooling tower water treatment programs (scale, corrosion, and Legionella control per ASHRAE 188), variable-speed pumping controls, and a building automation system capable of coordinating primary/secondary or variable-primary flow strategies. This isn't just a cost consideration — it's a staffing consideration. A building or campus without dedicated, trained operations staff (or a solid controls-integrated BAS with remote monitoring) will struggle to actually realize the part-load efficiency advantage chilled water plants are capable of on paper; a poorly operated chiller plant can easily underperform a well-selected DX system in practice.

Application-Driven Decision Criteria

In practice, the choice is rarely a pure economic optimization — it's driven by a handful of dominant factors specific to the project:

• Building size and campus configuration — single small-to-mid buildings favor DX; large single buildings or multi-building campuses favor central chilled water, since one plant can serve multiple structures and amortize its fixed cost across all of them.
• Load diversity — a campus or mixed-use building where different zones peak at different times benefits from a shared chilled water plant sized to the coincident (diversified) peak rather than the sum of individual zone peaks, which a fleet of independent DX units cannot capture.
• Redundancy requirements — mission-critical space (hospitals, data centers, labs) tends toward chilled water for centralized N+1/2N plant sparing.
• Available plant space and structural capacity — chilled water requires real central plant floor area, cooling tower roof or yard space, and pipe chase/shaft space that not every building, especially retrofits, has available.
• Refrigerant code exposure — dense-occupancy buildings using A2L-refrigerant DX equipment may hit ASHRAE 15 charge limits that push the decision toward chilled water regardless of scale.
• Phasing and future expansion — a chilled water plant can be designed with spare capacity (extra chiller taps, oversized piping mains) to absorb future building additions economically; adding DX capacity later means adding entirely new, independent equipment with no shared infrastructure benefit.

A useful rule of thumb many mechanical engineers apply during early conceptual design: default to DX below roughly 50,000–75,000 sq ft or under ~100 tons, default to chilled water above roughly 150 tons or for any multi-building campus, and treat the 100–150 ton range as a genuine trade study requiring an actual life-cycle cost analysis rather than a rule-of-thumb decision — first cost, energy cost at local utility rates, maintenance staffing availability, redundancy requirements, and refrigerant code exposure all need to be weighed explicitly for projects that fall in that middle zone.

Hybrid and Alternative Approaches Worth Knowing

The DX-vs-chilled-water framing is the classic comparison, but it's worth noting that VRF/VRV systems (covered in depth elsewhere on this site) occupy a middle ground — they're technically DX systems (refrigerant goes to the terminal unit) but with much better part-load modulation and zoning granularity than conventional single-stage DX, closing some of the part-load efficiency gap with chilled water while still confining the central compressor to one location, though refrigerant piping still runs throughout occupied floors, which keeps the ASHRAE 15 A2L consideration in play. Waterside economizers and airside economizers are also relevant to this decision independently of DX-vs-chilled-water — a chilled water plant with a waterside economizer (using cooling tower water to directly cool the chilled water loop, bypassing the chiller entirely) can achieve substantial "free cooling" hours in cool and dry climates, an efficiency lever that has no direct DX equivalent beyond airside economizer dampers.