- Singapore launched a liquid-cooling standard for AI data centres.
- Liquid cooling can cut data-centre energy use by over 30%.
Singapore has introduced what the Infocomm Media Development Authority (IMDA) and Enterprise Singapore (EnterpriseSG) describe as the world’s first liquid-cooling standard for tropical data centres. The standard addresses facilities supporting high-density computing equipment used for AI and other advanced workloads.
The Singapore Standard SS 726:2026, titled Liquid Cooling in Tropical Data Centres, was developed by IMDA and EnterpriseSG through the Singapore Standards Council. The standard covers the planning, design, deployment, operation, maintenance, and risk management of liquid-cooled systems.
Its foreword links the move toward liquid cooling with higher rack densities, noting that the data-centre industry is shifting from low-density to high-density racks with different infrastructure, space-planning, and operational requirements from conventional air-cooled systems. SS 726 also states that traditional air-cooling designs become less efficient as IT equipment increases in density and energy use. It builds on existing international liquid-cooling specifications while adding requirements and recommendations for systems operating under tropical temperatures and humidity.
AI racks push power higher
Newer AI systems illustrate the amount of power that can now be concentrated within a single rack. Nvidia’s DGX GB200 NVL72, for example, combines 72 Blackwell GPUs and 36 Grace CPUs in one rack-scale system and has an approximate rack power consumption of 120kW.
The system uses liquid cooling for its CPUs and GPUs, while components including networking and storage equipment remain air cooled. Liquid runs through manifolds and cold plates attached to the processors.
SS 726 addresses heat removal at the data-centre level, covering systems that use coolant to remove heat directly from components such as CPUs, GPUs, and tensor processing units. The standard does not cover rear-door systems that use liquid to remove heat from air behind equipment racks or other indirect liquid-cooling designs. Two-phase and hybrid single- and two-phase cooling systems are also outside its scope.
According to SS 726, removing heat directly from heat-generating components can maintain lower and more stable operating temperatures than air cooling at comparable rack densities. It can also reduce server-fan power consumption and the risk of thermal throttling, where processors reduce performance as operating temperatures rise. A greater share of heat removed by liquid can also support denser equipment deployments while reducing residual heat released into the data hall.
IMDA and EnterpriseSG said liquid cooling can reduce energy consumption at the data-centre level by more than 30% compared with traditional air cooling.
Cooling for tropical conditions
Singapore’s tropical climate adds specific requirements to data-centre cooling design. SS 726 identifies high humidity and ambient temperatures that often exceed 30°C as conditions that require additional operational measures.
Cooling already represents a large part of data-centre energy use. IMDA has previously estimated that cooling systems can account for up to 40% of total energy consumption in a typical data centre, with warmer tropical conditions requiring additional energy to operate cooling systems.
Singapore’s earlier Tropical Data Centre Standard allows facilities to operate safely at higher temperatures. IMDA said each 1°C increase in data-centre operating temperature can produce cooling-energy savings of around 2% to 5%.
SS 726 covers systems that remove heat directly from increasingly dense IT equipment. The standard states that liquid-cooling systems designed for fluid temperatures above 40°C are suited to tropical environments because they can use higher coolant temperatures for heat rejection.
This can reduce reliance on energy-intensive chillers, according to the standard. Higher-temperature liquid also carries waste heat at a higher temperature than conventional hot-air recapture, which can make the recovered heat more suitable for reuse where practical.
Tropical conditions also introduce specific operational risks. IMDA and EnterpriseSG said high temperatures and humidity can contribute to equipment corrosion, which can affect cooling efficiency and system reliability.
Installing liquid-cooled computing equipment requires supporting infrastructure beyond the processors themselves. SS 726 covers liquid-cooling facilities and equipment, pre-deployment testing and commissioning, operations and maintenance, risk management, and sustainability indicators.
The standard includes different coolant distribution unit, or CDU, configurations, including rack-mounted systems and floor-mounted liquid-to-liquid and liquid-to-air units. CDUs connect computing equipment with the wider cooling infrastructure and manage the flow of coolant through the system.
Liquid-cooled AI systems also require monitoring for coolant leaks and other cooling-system faults. Nvidia’s DGX GB systems, for example, incorporate leak detection to identify leaks before they damage equipment.
SS 726 provides guidance on cooling fluids, materials, maintenance, and other operational risks. It also supports the integration of liquid cooling with existing air-cooled data centres, covering facilities that use both cooling methods.
SS 726 also includes recommendations for measuring and optimising water use associated with liquid-cooling infrastructure.
IMDA and EnterpriseSG developed the standard with input from industry, academia, and technical experts.
“Liquid cooling is increasingly important to support the higher power densities required for AI and other advanced computing workloads, while improving cooling efficiency,” Aileen Chia, deputy chief executive at IMDA, said.
Singapore adds data-centre capacity
Singapore has also provisionally allocated 200MW of new data-centre capacity to four operators under its second Data Centre Call for Application. IMDA and the Singapore Economic Development Board announced the allocations on August 21, 2026.
The application exercise was launched in December 2025, with the agencies identifying AI and data-centre investment among its objectives.
Facilities selected under the programme must achieve a Power Usage Effectiveness (PUE) of 1.25 or better at full IT load. Applicants must also use equipment that meets or exceeds the energy-efficiency requirements in Singapore Standard SS 715:2025 and optimise equipment utilisation. At least 50% of the proposed new data-centre capacity must be powered through eligible green-energy pathways.
SS 715:2025 sets energy-efficiency requirements for data-centre IT equipment and requires covered equipment to operate safely at temperatures of up to 35°C. It sits alongside SS 697:2023, which covers operating data centres at higher temperatures in tropical conditions.
SS 726 extends that standards framework to liquid-cooling infrastructure, covering systems that remove heat directly from higher-density processors.
Singapore’s Energy Market Authority (EMA) expects peak electricity demand to increase by between 2.4% and 4.8% annually over the next decade, reaching between 9.6GW and 11.4GW by 2031. EMA identified energy-intensive sectors including data centres and semiconductor manufacturing among the contributors to demand growth.
The authority has also awarded Tuas Power the right to build 670MW of additional generation capacity, scheduled to enter operation by December 2031.
SS 726 complements Singapore’s Green Data Centre Roadmap, which IMDA introduced in 2024. The roadmap included plans to develop standards for IT equipment efficiency and liquid cooling.
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.
Tech Wire Asia is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
The post Singapore launches world’s first liquid-cooling standard for AI data centres appeared first on TechWire Asia.