Chiller vs VRF: What's the Difference?

Sit in enough design workshops and you notice that the chiller versus VRF debate rarely turns on engineering. It turns on who is in the room, and the operator who will live with the thing for two decades is usually not in it. So the argument is conducted in terms of first cost and plant footprint, and the differences that will actually matter in year eight, refrigerant handling, who is qualified to touch it, whether spares still exist, get discovered later. This guide is written from the operator's chair, because that is where the consequences land.

The message up front: there is exactly one root difference between these two schemes. A chilled water system moves cooling around the building as water. A VRF system moves it around as refrigerant. Nearly every other difference, plant space, riser size, structural weight, who can service it, what happens when it leaks, how it behaves at part load, how far it scales, follows from that one choice of medium. If you understand the medium, you can derive most of the comparison yourself.

1. The two schemes, briefly

A chilled water scheme is a central plant serving a hydronic distribution network. One or more chillers produce chilled water in a plant room or on the roof. Pumps push that water through insulated pipework, up risers, out along floor plates, to terminal units in the occupied space: fan coil units in small zones, air handling units where you need ventilation and larger air volumes, chilled beams or active terminals in some designs. The terminal unit passes room air over a water coil, the air gives up its heat to the water, and the warmed water returns to the chiller to be cooled again. The refrigerant in this scheme stays inside the chiller. It is confined to a machine in a plant room, and it never enters the occupied floors. The details of how the chiller itself produces that cold water are covered in the companion piece on what a chiller is and how chillers work, so I will not re-teach the vapour compression cycle here.

A VRF scheme, variable refrigerant flow, removes the water loop entirely. An outdoor condensing unit contains the compressors, and refrigerant pipework runs directly from that outdoor unit, up risers, through branch controllers or refrigerant distribution joints, to individual indoor units in each zone. The indoor unit contains a refrigerant coil and an expansion device. Room air passes over refrigerant, not water. The compressor modulates its output, usually with inverter driven compressors, to match the aggregate demand of all the indoor units connected to it. There is no chiller, no chilled water pump, no primary loop, and often no plant room at all. There is, instead, refrigerant pipework distributed throughout the building.

Both sit inside the same broader family of building services. If the vocabulary here is unfamiliar, the HVAC systems and components overview is the place to start before this comparison will make full sense.

2. The distribution medium is the root difference

This is the organising idea of the whole comparison, and once it clicks the rest becomes derivation rather than memorisation. Water is a benign medium. It is cheap, it is not regulated as a substance, a competent pipefitter can work on it, a leak is a housekeeping and damage problem rather than a safety problem, and you can isolate, drain and refill a branch without specialist certification. What water costs you is bulk. It is heavy, it needs large pipe cross sections relative to the energy it carries, it needs pumps, and pumps need power and space and maintenance. It also needs a machine at one end to chill it, and that machine needs a room.

Refrigerant is a dense and efficient medium. It carries heat as latent energy through phase change, which means it moves a great deal of energy through a small pipe. That is why VRF risers are slim, why the pipework is light, and why you can often thread a VRF scheme through a building that would never accept chilled water mains. What refrigerant costs you is that it is a controlled substance in a sealed system. Its containment matters, its leakage has safety and environmental consequences, the people who can legally and competently work on it are a narrower trade, and the system cannot simply be opened, drained and topped up by a general mechanical team.

3. Heat rejection, and where it happens

Both schemes must ultimately dump the building's heat outside. Where and how they do it is one of the more practical differences.

A chilled water scheme rejects heat either through an air cooled chiller, which has its own condenser coils and fans and sits outside, or through a water cooled chiller paired with a condenser water loop and a cooling tower. The water cooled route generally performs better thermodynamically, which is why larger schemes favour it, but it introduces an entire second wet system with its own dependencies: condenser pumps, water treatment, makeup water, blowdown, and the seasonal and hygiene regime that open evaporative equipment demands. That is not a small addition to an operations scope. The cooling tower explainer covers that dependency properly, and it is worth reading before committing to water cooled, because the tower is frequently the least well maintained part of a chilled water plant and the part most likely to embarrass you.

VRF rejects heat at the outdoor condensing unit, air cooled in most common configurations. That means no cooling tower, no condenser loop, no water treatment contract, and no evaporative hygiene regime. It also means the heat rejection is distributed: rather than one plant location, you have outdoor units wherever the vertical and horizontal pipe runs allow, typically on roofs, terraces, plant decks and sometimes screened balconies. Distributed heat rejection is easier to install and harder to manage. You now have more locations to access, more units in exposed positions, more coils to keep clean, and a real risk of short cycling hot air between units if the architectural screening was designed for appearance rather than airflow. On plenty of roofs VRF units are boxed in so tightly that they are recirculating their own discharge, and the building is quietly paying for that every summer.

4. Simultaneous heating and cooling: VRF's genuine advantage

This is the one area where I think VRF has a real and often decisive edge, and it is frequently undersold by people arguing the case on footprint alone.

Buildings are not uniform. At the same hour on the same day, a south facing meeting room can be rejecting heat while an interior office on the shaded side is calling for warmth, a data closet is rejecting heat all year, and a lobby is fighting infiltration. A heat recovery VRF configuration, usually a three pipe arrangement or a two pipe arrangement with branch controllers, can serve both demands at once and, crucially, can move the heat it extracts from the zone that does not want it into the zone that does. The heat rejected by the meeting room becomes the heat delivered to the interior office, and the outdoor unit only has to make up the difference.

A chilled water scheme can do simultaneous heating and cooling too, but it does it differently and with more infrastructure. You either add a heating source and a separate heating water distribution, giving you a four pipe scheme with two complete loops and terminal units with two coils, or you use reversible heat pump chillers with the associated hydronic complexity, or you accept a changeover arrangement and the comfort compromise that comes with it. All of those are viable. None of them is as neat as a heat recovery VRF doing it inherently as part of how the refrigerant circuit works.

So if the building has a genuinely mixed simultaneous load profile, deep floor plates with perimeter and core zones behaving differently, high internal gains in some zones and not others, this is the strongest single argument for VRF and it should be weighed seriously. It is also a claim worth testing rather than assuming. Plenty of buildings are pitched as needing heat recovery when in practice the whole floor plate is calling for cooling almost every occupied hour, in which case you have paid for a capability you will rarely use.

5. Plant space, riser space, and the structural consequence

This is where VRF wins the arguments it usually wins, and the reasons are sound even if they are overstated.

A chilled water scheme needs a plant room, and a real one: space for the chillers themselves, clearance for tube pull on shell and tube machines, pump sets, buffer or thermal storage vessels if the design calls for them, expansion and pressurisation equipment, water treatment, electrical switchgear, and enough circulation space for a technician to work safely and for a compressor or bundle to be removed one day. It needs vertical riser shafts sized for insulated flow and return mains. And it needs the structure to carry all of that: a chiller is a heavy machine, a full water loop is a large mass of water in the ceiling voids and shafts, and the plant room floor loading is a real design input rather than an afterthought.

A VRF scheme, by comparison, is remarkably light on all three counts. Indoor units are small and can often sit in a ceiling void that would never accept a fan coil unit plus insulated water pipework plus a condensate route plus valve access. Refrigerant risers are slim. There is no plant room in the traditional sense, only outdoor unit space and branch controller access positions. Structurally the whole scheme is lighter, and in a building where the frame is already designed, or is historic and cannot be strengthened, that difference is not a nice to have, it decides the project.

6. Commissioning and controls

Commissioning a chilled water scheme is a hydraulic exercise with a long and well understood body of practice behind it. You flush and fill, you prove water quality, you set up the pumps, you balance the network so that every terminal gets its design flow, you prove valve authority and control response, and you demonstrate that the plant stages and sequences correctly. It is laborious, it is frequently rushed, and unbalanced water is one of the most common chronic faults in commercial buildings. But it is transparent: flow, pressure and temperature are all measurable with ordinary instruments, and a competent commissioning engineer can diagnose the system from the outside.

Commissioning VRF is a different animal. Pipe runs must be within the manufacturer's permitted lengths and height differences, joints must be correctly made, the system must be properly evacuated and charged to the calculated quantity for the actual installed pipe volume, and the whole thing is then commissioned substantially through the manufacturer's own commissioning tool or service software, which reads the system's internal parameters and confirms addressing and configuration. The diagnostics are excellent when you have the tool and the training. They are largely opaque when you do not.

On controls, VRF arrives with a capable proprietary controller and, typically, an optional gateway to expose points to a building management system. Chilled water schemes are more naturally integrated: valves, sensors, pumps and chiller control panels are more often open protocol or at least conventionally interfaced, and the sequences are written by the controls contractor rather than embedded in the equipment. If BMS integration and open data access matter to you, and for anyone running analytics or fault detection they should, this is a real consideration. A VRF scheme behind a limited gateway can be a frustrating source of data compared with a chilled water plant where you can read every temperature, flow and valve position. That data access question is exactly what determines whether plant level analytics are even possible, a theme I take up in chiller plant analytics.

7. Part load behaviour, qualitatively

I am deliberately not going to give you efficiency figures here, because published performance numbers depend entirely on the test condition, the rating method and the specific product, and quoting them in a general comparison misleads more than it informs. What is worth understanding is the shape of the behaviour.

Buildings spend most of their hours well below design load, so part load behaviour, not peak performance, determines the annual outcome for both schemes. VRF is inherently good at part load in one specific sense: the compressor modulates continuously to match the summed demand of the connected indoor units, and each zone is controlled independently at the point of use. A floor with three occupied offices out of twenty is genuinely cooling three offices. There is no distribution medium being circulated at full flow to serve a small load. That granularity is the real efficiency story with VRF, and it is strongest in buildings with diverse, intermittent, zone by zone occupancy.

A chilled water scheme achieves good part load performance through design rather than inherently: multiple chillers staged so that the running machines operate in their efficient band, variable speed pumping so flow follows load, variable speed compressors or unloading arrangements on the machines themselves, and terminal control that actually closes valves when zones are satisfied. Get that design and its commissioning right and a chilled water plant performs very well across the load range. Get it wrong, and the classic failure appears: everything runs, the temperature difference across the plant collapses, flow rises to compensate, pumps work hard, and the plant is busy without doing useful work. Whether a chilled water plant is efficient at part load is therefore substantially a question about design and operation rather than about the technology.

On the VRF side the equivalent qualitative caution is that performance depends heavily on the installation being correct: pipe lengths within limits, correct charge, clean coils, and outdoor units that are not recirculating their own discharge air. A poorly installed VRF scheme does not announce itself the way an unbalanced water system does. It just quietly underperforms.

8. The refrigerant question, which is the one that gets glossed

This is one of the most consequential differences and the one most often skipped in a comparison, so I want to be direct about it and equally careful not to overstate what I can verify.

In a chilled water scheme the refrigerant charge is confined to the chiller, in a plant room or on a roof, in a machine designed and located for the purpose. The pipework that runs through the occupied building contains water. In a VRF scheme the refrigerant is the distribution medium, which means refrigerant pipework runs through risers, ceiling voids, corridors and occupied rooms, and the refrigerant coil sits inside the indoor unit in the space.

That difference has design consequences that have nothing to do with efficiency or cost. Distributing refrigerant through occupied areas raises questions about leak detection, about the ventilation and room volume relationship for the spaces served, about what happens to a leak in a small enclosed room, about alarm and shutdown provisions, and about how you demonstrate all of that to a reviewing authority. A chilled water scheme raises those questions only for the plant room, which is a single, controlled, purpose designed space. It does not raise them for every room in the building.

I am not going to tell you which refrigerants are permitted, restricted or on the way out, what their environmental metrics are, what charge is allowable in a given room volume, or which document governs it. Those rules are genuinely regulated, they differ by jurisdiction, and they have been changing. What I will tell you plainly is that this is a designer and regulator question, not a procurement question: refrigerant selection and leak related safety design for a VRF scheme must be resolved by a qualified designer against the rules current in your jurisdiction at the time of design, and confirmed with the authority that will review it. Anyone who answers this question for you from memory, including a vendor, is guessing.

9. Serviceability, who can work on it, and spares over twenty years

This section is where the operator's view diverges most sharply from the designer's, and it is the one I would most want a client to read twice.

A chilled water plant is largely a mechanical and electrical estate. Pumps, valves, strainers, pipework, insulation, coils, drives, sensors and actuators are all serviceable by general mechanical trades that any decent FM contractor already employs. The chiller itself is the specialist item, and it is one item, in one room, on a maintenance contract. Components are broadly interchangeable across manufacturers: a pump is a pump, a two port valve is a two port valve, a fan coil unit motor can be sourced. If your chiller manufacturer becomes difficult twelve years in, you replace a machine and keep the entire distribution network. That decoupling of plant from distribution is a genuinely valuable property over a long asset life. The routine regime for that plant is covered in boiler and chiller preventive maintenance, and the wider estate schedule in preventive maintenance for HVAC systems.

VRF is a different service proposition. Because refrigerant is the distribution medium, work on the distribution system is refrigerant work, which requires refrigerant qualified technicians with the proper handling equipment and, in most jurisdictions, the appropriate certification and record keeping. Diagnostics are usually performed with the manufacturer's service tool reading the system's own parameters, which means effective fault finding is tied to having that tool, that software and that training. Indoor units, branch controllers, boards and sensors are manufacturer specific parts, not commodity items.

Over a fifteen to twenty year life that creates a lock in worth thinking about honestly. You are committed to one manufacturer's ecosystem for spares, for control boards, for gateway firmware and for the service tool, and to a service market that may be thin in your locality. This is not an argument against VRF, and major manufacturers generally support their systems well. It is an argument for asking specific questions before you commit: who locally is authorised and equipped to service this, how many of them are there, what is the spares availability commitment, what happens to the control boards when this product generation is superseded, and can our own FM team be trained and tooled or are we permanently dependent on one service provider.

10. Retrofit suitability, and what happens as load grows

Two situational factors pull hard in opposite directions, and between them they decide a large share of real projects.

Retrofit favours VRF, often decisively. In an occupied building, a heritage building, a phased refurbishment, or any structure where you cannot create a plant room, cannot add significant structural load and cannot find riser space for insulated water mains, VRF is frequently the only scheme that can actually be installed. Slim refrigerant pipework threads where water mains cannot. Work can be sequenced floor by floor, or tenancy by tenancy, with a new outdoor unit serving each phase, so the building keeps operating and the capital is staged. The absence of a wet distribution system inside a historic fabric is itself a risk reduction. If someone tells you VRF was chosen for a refurbishment on installability grounds, that is usually a sound reason rather than a lazy one.

Scale favours chilled water, and increasingly so as load grows. As the building gets larger, the load gets bigger and the distribution distances get longer, the central scheme's advantages compound. Large central machines and staged plant handle large aggregate loads in a way that multiplying VRF systems does not match elegantly. Water can be pumped further than refrigerant can practically be run, and refrigerant pipe length and height limits constrain how VRF can be laid out in a tall or deep building. Where large ventilation air volumes must be conditioned, a chilled water coil in an air handling unit is the natural solution, and zone level trim through fan coil units completes a scheme that is well understood by every commissioning engineer and FM contractor you will ever hire. A central plant also gives you a single connection point for thermal storage, for future plant replacement, or for a district cooling supply.

I deliberately will not give you a capacity number as a switchover rule, because any such threshold is a function of building geometry, load diversity, available space and local market, not a property of the technologies. What is reliable is the direction: small to medium, tenancy scale, retrofit, diverse intermittent occupancy points to VRF; large, dense, ventilation heavy, long distribution, single ownership points to chilled water. The interesting projects are in the middle, and those are decided on the specifics.

11. The comparison in one table

Everything above, compressed. Read it as tendencies rather than absolutes, because a well designed example of either scheme beats a poorly designed example of the other on nearly every row.

12. Deciding: situations and the likely better fit

Design decisions are made about specific buildings, not categories, so here is the same comparison expressed as situations. Treat the right hand column as where the burden of proof starts, not as a verdict: a local constraint or an unusual load profile can reverse any row here, and often should.

The idea to walk away with

There is no winner here, and the comparison is not close to being settled by technology. Chilled water and VRF are two different answers to the question of how you move cooling from where you make it to where you need it, and each answer is correct in the circumstances that suit it. Chilled water is the right answer when the building is large, dense, ventilation led, centrally owned and expected to outlive several generations of plant, and when the organisation running it wants a widely serviceable estate with no controlled substance in the occupied floors. VRF is the right answer when the building is constrained, occupied, historic, phased, tenanted, or has a genuinely mixed simultaneous load profile, and when the organisation can secure competent refrigerant service for the asset's whole life.

What I would resist is the framing that one of them is modern and the other is legacy. That framing is a sales device, not an engineering judgement. Both are current, both are actively developed, and both are installed in new buildings every week by serious engineers. The question is never which is better. It is which set of trade offs this building, this owner and this operations team can live with for twenty years. And if you take only one thing away, take the medium. Water or refrigerant. Ask what the building distributes, and you can reason your way to almost every other difference without needing anyone's brochure.

Final thoughts

The pattern that recurs is that this decision gets made on two inputs, first cost and plant footprint, and reviewed against a third, energy modelling, while the inputs that dominate the operational decades go unexamined. Who will service this. How many of them exist within an hour of the site. What the refrigerant safety case requires of the facilities team after handover. Whether the control system will give up its data or hold it behind a gateway. Whether the spares for these indoor units will still be available in year fourteen. None of those questions are hard to ask. They are simply not on the agenda when the decision is taken, because the people who will answer for them are not in the room.

So my practical advice is procedural rather than technical. Put the operator in the room while the choice is still open. Ask the design team to document the refrigerant safety case and the BMS data strategy as deliverables rather than assumptions. Test the simultaneous load claim against expected occupancy before paying for heat recovery. Survey the local service market honestly before signing. Do that and you will very likely still choose the scheme the footprint argument pointed you to, but you will choose it knowing what you have taken on, which is the difference between a decision and a default.

Related reading: What is a chiller: types and how chillers work, HVAC: meaning, systems and components, What is a cooling tower and how it works, AHU explained, FCU explained, Boiler and chiller preventive maintenance, Chiller plant analytics, Preventive maintenance for HVAC systems.

Further reading from the professional bodies: ASHRAE and CIBSE .