Singapore’s Tropical Data Centre Testbed 2.0: Practical Lessons for AI-Ready Facilities

How Singapore facility teams can connect cooling, power, controls, water and operational data before higher-density computing arrives.

Higher-density computing is changing the engineering conversation for data centres and other digital facilities. Cooling capacity, electrical demand, control logic, water use and operational data can no longer be treated as separate topics.

Singapore’s Sustainable Tropical Data Centre Testbed 2.0 provides a timely local reference point. Hosted by the National University of Singapore’s College of Design and Engineering, the programme brings research and industry partners together to evaluate technologies for data-centre operations in Singapore’s hot and humid environment.

Recent announcements from Johnson Controls and NUS describe work involving thermal management, smart controls, advanced liquid cooling, grid-responsive operations and lower-resource infrastructure. These activities are intended to test and validate technologies under local conditions. They should not be interpreted as a single commercial solution suitable for every facility.

For Singapore facility managers, building owners, warehouse operators and SMEs, the more useful question is: what practical preparation can begin now?

1. Treat cooling as part of a wider operating system

Traditional ACMV maintenance often focuses on equipment condition, setpoints, comfort and energy consumption. These remain important, but higher-density digital infrastructure requires a broader view.

Cooling performance depends on the relationship between IT or process load, air distribution, chilled-water or refrigerant systems, pumps, fans, controls and the building envelope. A change in one area can affect the others. For example, increasing airflow may support equipment temperatures but also increase fan energy. Lowering a supply-air temperature may provide additional margin but increase cooling demand.

Facility teams should develop a clear operating picture that links:

• Computing, process or equipment load;
• Room and rack-level temperatures where relevant;
• Supply and return air conditions;
• Chilled-water temperatures, flow and differential pressure;
• Fan, pump and compressor speed;
• Power consumption by major system; and
• Alarm, maintenance and control-system events.

This does not require every organisation to install a complex platform immediately. It does require reliable measurement and a consistent way to relate cooling demand to actual load.

2. Improve sensor quality before adding more automation

Automation is only as useful as the data supporting it. Inaccurate sensors, inconsistent naming, missing trends and poorly positioned measurement points can lead to inefficient or unsafe control decisions.

Before introducing advanced analytics or AI automation, engineering teams should review the basics:

• Are temperature, humidity, pressure, flow and power sensors calibrated and maintained?
• Are readings time-synchronised across BMS, EPMS, DCIM or other systems?
• Can operators distinguish measured values from estimated or calculated values?
• Are alarms prioritised according to operational risk?
• Is historical data retained long enough to identify seasonal and load-related patterns?

For warehouses and facilities with temperature-sensitive operations, the same principle applies. Real-time monitoring can be valuable, but it must be connected to clear response procedures. A dashboard that shows a temperature deviation without defining who responds, how quickly and with what authority is not a complete control strategy.

3. Prepare for liquid cooling without assuming it is the answer everywhere

Advanced liquid cooling is one of the technologies being considered in the testbed’s AI-ready focus. Liquid cooling may support higher-density equipment, but its suitability depends on the load profile, equipment design, water or fluid management approach, maintenance capability, leak detection, redundancy and integration with the wider heat-rejection system.

Facility managers should therefore treat liquid cooling as an engineering option to be evaluated, not a universal replacement for air cooling. Practical questions include:

• What equipment densities and temperature limits are expected?
• Where will heat be transferred and rejected?
• How will leaks, fluid quality and maintenance access be managed?
• Can the system operate safely during partial equipment or control failure?
• What changes are required to power, space planning, drainage and emergency response?

Testing under Singapore conditions is particularly relevant because tropical heat and humidity influence heat rejection, condensation risk, maintenance conditions and overall system performance.

4. Connect power planning with cooling planning

More computing capacity generally means more electrical demand and more heat to remove. A cooling upgrade planned without a corresponding power review can create constraints elsewhere in the facility.

Engineering teams should review the relationship between incoming supply, distribution boards, UPS systems, generators, cooling plant, pumps, fans and future equipment loads. The review should consider normal operation, staged expansion, maintenance conditions and failure scenarios.

The NUS announcement on the testbed also refers to grid-responsive operations. For facility owners, this points to a wider opportunity: energy systems may need to respond more intelligently to changing loads, tariffs, grid conditions or operational priorities. Any automated response should be designed with clear limits, override controls and protection for critical loads.

5. Use controls coordination instead of isolated optimisation

Energy optimisation often fails when each subsystem is tuned independently. A fan-control strategy may conflict with air-conditioning control. A chilled-water reset may reduce efficiency at one load condition but reduce resilience at another. An occupancy or equipment schedule may not match the actual thermal profile.

Singapore Green Building Council content on coordinated fan and air-conditioning control illustrates the value of looking at these systems together. The exact benefit will vary by building, equipment, controls and operating conditions, so published potential savings should not be assumed for every site.

A practical controls review should examine:

• Setpoint hierarchy and control ownership;
• Deadbands, staging and sequencing;
• Minimum operating limits for critical equipment;
• Manual override and fallback modes;
• Interaction between ACMV, electrical and monitoring systems; and
• Whether control changes are documented and verified after implementation.

6. Make water use part of the engineering decision

Cooling efficiency is not measured only in kilowatt-hours. Water consumption, treatment requirements, discharge arrangements and maintenance resources can also affect sustainability and operating risk.

Teams considering cooling-tower optimisation, evaporative methods or liquid-cooling systems should establish a site-level water balance and identify the operational trade-offs. A solution that reduces electricity use but introduces difficult water-management requirements may not be appropriate for every facility.

For building owners and SMEs, the first step may simply be to monitor water consumption associated with cooling and investigate abnormal changes. For larger digital facilities, water data should be reviewed alongside energy, thermal and maintenance data.

7. Validate technology before full deployment

The Sustainable Tropical Data Centre Testbed 2.0 reinforces the value of testing emerging solutions in local operating conditions. Singapore’s climate, land constraints, energy infrastructure and operational requirements may produce different results from overseas demonstrations.

Facility teams can apply the same principle at a smaller scale through a controlled pilot. Define the baseline, select measurable performance indicators, establish operating limits and document the conditions under which the technology is tested. Where possible, compare energy, temperature stability, water use, alarms, maintenance effort and operator workload before and after the pilot.

BCA’s Innovation Sandbox is another example of Singapore’s broader approach to testing and demonstrating emerging technologies before wider adoption. The lesson for private facilities is straightforward: verify performance in context, rather than relying only on brochures or laboratory results.

From testbed lessons to practical facility readiness

AI-ready infrastructure is not created by installing one new cooling product or one analytics dashboard. It depends on a connected operating model covering thermal management, electrical capacity, control logic, water use, monitoring quality and human response.

Singapore facility teams can begin with a readiness review:

1. Map current cooling, power, water and controls systems.
2. Check the quality and availability of operational data.
3. Identify thermal, electrical and control bottlenecks.
4. Define critical operating limits and fallback modes.
5. Assess whether a targeted pilot is appropriate.
6. Set measurable outcomes before investing in new technology.

ISS supports businesses seeking to connect engineering operations, facility management and AI automation in a practical way. Contact ISS to discuss your engineering, facility management or AI automation requirements.

Reference points: Johnson Controls announcement on Sustainable Tropical Data Centre Testbed 2.0; NUS announcement on EdgeConneX joining the testbed; Singapore MDDI digital infrastructure policy update; BCA Innovation Sandbox.