Most large buildings need cooling. They do not all need to manufacture that cooling independently.
A conventional commercial building has its own chiller plant, pumps, cooling towers or heat-rejection equipment, maintenance team and spare capacity for hot days or equipment failure.
District cooling moves much of that machinery out of individual buildings and into a shared utility network.
Large central plants produce chilled water efficiently. Underground supply pipes send that chilled water toward connected buildings. Heat exchangers inside each building transfer heat from the building’s internal air-conditioning loop into the district water. The warmer return water then flows back to the district plant to be cooled again.
SP Group’s Marina Bay system has now operated for 20 years. As of August 2026, it serves 28 buildings through about 5 kilometres of piping, avoids about 25,000 tonnes of carbon emissions annually and has recorded zero cooling-service disruptions over the full 20-year operating period.
The operating loop is: central chillers remove heat → chilled water enters district supply pipe → building heat exchanger absorbs indoor heat → building air-conditioning loop cools rooms → warmed district water returns → central plant rejects the heat → water is chilled again.
Quick answer: why can one shared system be more efficient?
- Larger chillers can operate more efficiently: central plants can use high-performance equipment and expert optimisation.
- Diversity reduces duplicated spare capacity: not every building reaches maximum cooling demand at exactly the same moment.
- Maintenance is concentrated: specialist teams operate fewer, larger plants rather than thousands of independent small ones.
- Buildings gain usable space: connecting can remove the need for large on-site chiller plants.
- Redundancy becomes shared: several chillers and plants can support one network instead of every building carrying its own full backup set.
1. A chiller is a heat-moving machine
Air-conditioning does not manufacture cold as a substance.
A chiller uses a refrigeration cycle to move heat from chilled water into another heat-rejection medium.
The chilled water then travels to cooling coils or heat exchangers where it absorbs heat from occupied spaces.
District cooling keeps that basic thermodynamics and changes who owns the largest equipment.
2. The district plant becomes a cooling utility
Instead of Building A, Building B and Building C each owning a separate chiller plant, the district operator produces chilled water for all of them.
The model resembles electricity supply: the customer buys a useful energy service from a shared network rather than operating the full upstream production system independently.
The building still needs internal air-handling and chilled-water equipment, but the largest cooling-production machinery can sit elsewhere.
3. Supply and return pipes create the district loop
District cooling normally uses a pair of insulated pipes.
The supply pipe carries colder water from the plant toward customers. The return pipe carries warmer water back after it has absorbed heat.
The water therefore circulates rather than being consumed like drinking water.
The network is a heat-transport system more than a water-supply system.
4. The building heat exchanger creates an ownership boundary
The district operator does not need to send its chilled water through every fan coil inside the building.
A heat exchanger transfers thermal energy between the district loop and the building’s internal cooling loop without necessarily mixing the two water circuits directly.
This creates a clean operational boundary: the district utility supplies cooling energy to the interface; the building distributes that cooling internally.
5. Diversity is the hidden efficiency gain
One office tower may peak in the afternoon.
A hotel may have a different load profile. A retail complex may remain busy later into the evening.
If each building sizes equipment for its own maximum demand, the district contains a great deal of duplicated capacity that is rarely used simultaneously.
A shared system can size against the combined diversified peak instead of the mathematical sum of every building’s separate maximum.
6. Large plants can run chillers nearer their efficient operating range
Chiller efficiency changes with load and operating conditions.
A central plant with several machines can decide which chillers should run and at what loading rather than forcing one building’s machine to operate inefficiently at very low load.
The larger fleet creates more combinations for matching plant output to actual district demand.
7. Shared redundancy uses fewer idle machines overall
Every critical chiller plant needs backup capacity for maintenance and equipment failure.
If 20 buildings each carry a full independent standby machine, the district can accumulate large amounts of rarely used plant.
A district system can provide resilience across a shared fleet of chillers, plants and network routes while still maintaining required service reliability.
Resilience becomes pooled rather than duplicated.
8. Marina Bay began commercial operations in 2006
SP’s Marina Bay district cooling system began commercial operations in May 2006, with One Raffles Quay among the first developments served.
The network later expanded to support major developments including Marina Bay Sands, Marina Bay Financial Centre and other central-area buildings.
The system has therefore grown together with the district rather than appearing after every building had already installed permanent standalone chillers.
9. Common Services Tunnels made utility expansion easier
Marina Bay was planned with district-scale infrastructure including Common Services Tunnels.
These underground corridors make it easier to route and maintain utilities without repeatedly excavating major streets.
District cooling therefore benefited from urban planning that reserved space for shared utilities before the entire precinct was built out.
10. The network now serves 28 buildings through about five kilometres of pipes
SP’s August 2026 update states that the Marina Bay district cooling network currently serves 28 buildings through a five-kilometre piping network.
The system has also expanded beyond the original financial-district core toward Raffles Place and nearby central precincts.
The cooling network therefore behaves like other utilities: once the main backbone exists, nearby developments can become increasingly practical to connect.
11. Satellite plants extend capacity without rebuilding the original plant endlessly
As the district grows, additional cooling capacity can be added at satellite locations connected to the same pipe network.
SP is developing a new satellite chiller plant at the George Street electricity substation, expected around 2028.
The plant will be co-located with electricity infrastructure, using another form of land stacking in the dense CBD.
The network can therefore gain capacity near new demand without requiring every new building to create its own plant room.
12. New Havelock connections show district cooling moving beyond Marina Bay itself
SP’s 2026 expansion plan includes new connections in the Anson, Enggor and Havelock areas.
The Havelock cluster will include a district cooling plant within Union Square Central linked to the future George Street satellite plant.
What began as one planned precinct is becoming a larger central-area cooling network with multiple production nodes.
13. Buildings can recover valuable floor area
Standalone chillers, pumps and associated plant rooms occupy expensive space.
When OUE Bayfront connected to the district network in May 2026, SP stated that more than 22,600 square feet of former on-site chiller-plant space could be repurposed.
The value of district cooling therefore is partly thermodynamic and partly real-estate economics.
14. Connection can reduce total cost of ownership
SP’s 2026 Marina Bay update states that connection to the network can reduce total cooling cost of ownership by up to 15% for participating developments under the relevant comparison.
The saving can come from lower upfront equipment requirements, shared operations and maintenance and more efficient central plant utilisation.
The building buys cooling as a utility rather than owning every upstream asset required to make it.
15. District cooling still depends on electricity
Central chillers and pumps consume substantial electrical power.
District cooling therefore does not remove the building sector from the electricity grid.
Its advantage is using electricity more efficiently to deliver the same cooling service and operating the plant at district scale.
The cooling network and power network are separate utilities whose performance is tightly coupled.
16. Thermal storage can separate cooling production from cooling demand
District systems can use thermal energy storage where designed to produce chilled water during periods of lower demand and release stored cooling later.
This shifts part of chiller electricity demand through time without using electrochemical batteries.
The water or other thermal medium stores a temperature difference rather than electrical charge.
17. One shared network creates a larger common-mode risk
Integration creates efficiency and also concentrates dependency.
If one small standalone building chiller fails, one building suffers. If a district backbone failed completely, several customers could be affected.
This is why redundancy in plants, pumps, pipes, controls and operating procedures is central to district cooling design.
SP reports zero service disruptions over Marina Bay’s first 20 years, illustrating how shared infrastructure can be highly reliable when resilience is designed into the network.
18. A worked example: office tower joins an existing district network
Imagine a new office tower planned beside an established district cooling main.
Instead of building a full standalone central chiller plant, the development installs its district-cooling interface, internal chilled-water circulation and air-handling equipment. The utility pipe supplies chilled water to the heat exchanger. The building removes heat from occupied spaces and transfers that heat into the district return water. The warmed water returns to the central network and is cooled again at the most efficient available district plant.
The building still owns comfort. The district owns much of the cooling factory.
19. Common misconceptions
Misconception: District cooling pumps cold air through underground tunnels.
No. the district network mainly distributes chilled water; buildings use that cooling energy in their own internal systems.
Misconception: Buildings connected to district cooling need no air-conditioning equipment.
No. they still need heat exchangers, pumps, air-handling units and internal distribution; they avoid much of the central cooling-generation plant.
Misconception: One central system is efficient only because the chillers are larger.
No. diversity, shared redundancy, expert operation and network-scale optimisation are major parts of the benefit.
Misconception: District cooling is independent of the electricity grid.
No. chillers and pumps remain major electrical loads.
Misconception: Shared infrastructure is automatically more fragile.
It can concentrate dependency, but engineered redundancy and professional operation can produce very high service reliability.
20. The deeper idea: district cooling turns spare capacity into a shared resource
Every standalone building protects itself by owning enough chiller capacity for its own hot afternoon and its own equipment failure.
At district scale, those separate safety margins overlap.
One network can pool cooling demand, backup equipment, maintenance expertise and plant efficiency across many buildings.
The system works because the city stops asking every tower to solve the same thermodynamic problem independently.