Singapore faces hot, humid conditions for more than 85% of the year; it has introduced a liquid-cooling standard for data centres

Singapore has introduced a national standard designed to help data centres adopt liquid cooling in the tropics, as rising computing demands make it harder to rely on conventional air-based systems alone. Singapore Standard SS 726:2026, Liquid Cooling in Tropical Data Centres, was jointly developed by the Infocomm Media Development Authority and Enterprise Singapore through the Singapore Standards Council, with input from data centre operators, technology providers, academics and technical experts. The Singapore government describes it as the world's first standard specifically aimed at helping data centres adopt liquid cooling technologies in tropical climates. The standard addresses system design, commissioning, operation, risk management and integration with existing air-cooled facilities, while also including guidance on energy and water efficiency.Singapore spends most of the year outside the climate window for direct air coolingThe challenge becomes clearer when Singapore's climate is compared with the conditions needed for direct air-side economisation, commonly called air-free cooling. In a study published in Scientific Reports titled Limitations to air free cooling in data centers under rising heat and humidity, researchers led by Christina Karamperidou of the University of Hawai‘i at Mānoa analysed hourly climate data and data centre locations across major markets. They defined air-free-cooling-limiting conditions as inlet temperatures above 27°C occurring alongside either a dew point above 15°C or relative humidity above 70%. Singapore had the highest exposure among the major markets studied, with such conditions occurring during more than 85% of the year over the study's historical period. The Pearl River Delta, encompassing Guangzhou, Shenzhen and Hong Kong, exceeded the thresholds for more than 40% of the year. The researchers stress that these figures represent climate conditions that constrain direct air cooling, rather than proving that compressor-free cooling is impossible at every site or under every system configuration.AI is making the heat problem harderAt the same time, the computing workloads being placed inside data centres are becoming substantially more heat-intensive. Singapore's SS 726 factsheet says most data centres in the country currently use air cooling, but AI applications increasingly rely on accelerators such as graphics processing units that consume more power and generate more heat than conventional servers. An industry estimate cited by the Singapore government puts the power consumption of an AI server rack at up to 50 times that of a conventional rack used for general-purpose computing. Liquid cooling brings coolant much closer to the heat-generating hardware, allowing heat to be transferred more efficiently than with air alone. Singapore says deploying liquid cooling can reduce data centre energy consumption by more than 30% compared with conventional air cooling, although the actual savings depend on system design and operating conditions.Tropical humidity creates problems beyond simply removing heatMoving from air to liquid cooling, however, introduces another set of engineering challenges in a tropical environment. Singapore's factsheet identifies equipment corrosion, microbial growth and biofilm formation in cooling fluid as risks associated with the country's high temperature and humidity. Liquid systems also depend on pumps, piping and heat exchangers, while leaks can cause substantial damage when coolant reaches sensitive IT equipment. Integrating liquid cooling into an existing air-cooled data centre can be complex and expensive because the cooling infrastructure becomes more tightly connected to the server racks. SS 726 therefore incorporates guidance on coolant distribution units, piping materials, fluid quality, floor loading, monitoring and risk management, alongside practices adapted from the Open Compute Project and the American Society of Heating, Refrigerating and Air-Conditioning Engineers.The standard covers the technology from installation to daily operationSS 726 is intended to provide guidance across the full lifecycle of a liquid-cooled data centre, beginning with design and pre-deployment and continuing through commissioning and day-to-day operation. Its recommendations cover cooling-loop and equipment design, piping-material selection, fluid-quality requirements, monitoring and management, as well as ways to integrate liquid cooling with existing air-conditioning infrastructure. The standard also incorporates input from Singapore's national water agency, PUB, on measuring and calculating water utilisation efficiency in liquid-cooled facilities. Industry participants including AirTrunk, Meta, the Open Compute Project Foundation and ST Telemedia Global Data Centres contributed to the working group, with ST Telemedia saying it plans to use the standard to guide future liquid-cooling deployments across its portfolio.Climate change could make the cooling window even narrowerThe need for such guidance could extend beyond Singapore as temperatures and humidity rise elsewhere. The Scientific Reports study analysed conditions from 1980 to 2024 and projected changes through 2050. It found that Singapore, already exposed to air-free-cooling-limiting conditions for more than 85% of the year, is projected to exceed the study's temperature and humidity thresholds year-round under its future modelling. Markets that currently experience less exposure, including Dublin, Chicago and Northern Virginia, were also projected to see substantial increases in threshold-exceedance hours. The researchers emphasise that their analysis does not prescribe a particular cooling technology or data-centre location. Instead, it shows how a warming and increasingly humid climate could reduce the amount of time that direct air-side economisation can operate within recommended environmental conditions, increasing the importance of alternative and hybrid cooling approaches.You use AI every day. Now get your AI Quotient. Take the AIQ test.