Southeast Asia’s AI Data Boom Strains Tropical Power Grids

Southeast Asia’s AI Data Boom Strains Tropical Power Grids

The Challenge of Cooling Data Centres in the Tropics

Singapore, located just one degree north of the equator, experiences a climate where temperatures rarely drop below 25 degrees Celsius and humidity remains high throughout the year. This makes it an unfavourable location for cooling data centres, which typically require operating temperatures between 18 and 27 degrees Celsius. Despite these challenges, Singapore has become a hub for data-centre infrastructure, with over 70 facilities and more than 1.4 gigawatts of capacity.

The city-state’s dense concentration of data centres reflects its ambitions in artificial intelligence and the growing demand for digital infrastructure across Southeast Asia. However, the energy consumption of these facilities is rising rapidly, with the International Energy Agency estimating that data-centre electricity demand could double by 2030 if current trends continue. In some markets, this could account for as much as 30 per cent of total national electricity demand, raising concerns about the strain on power grids.

Engineering Challenges in the Tropics

Professor Lee Poh Seng, head of the National University of Singapore’s mechanical engineering department, emphasizes that while Southeast Asia can host large-scale digital infrastructure, the tropical climate requires careful attention to efficiency. High ambient temperatures and persistent humidity limit the use of “free cooling,” which involves using outside air to manage heat. AI has further complicated matters, as modern data centres must manage extreme localised heat fluxes at the chip, server, and rack level.

“The engineering has to be much tighter and the controls have to be smarter,” Lee said. “That is especially true as the region expands cloud and AI capacity.”

Innovations in Cooling and Efficiency

In response to these challenges, operators are implementing new strategies to improve efficiency. For example, BDx Data Centres recently raised operating temperatures from 23 to 25 degrees Celsius at its Paya Lebar facility, resulting in a 7 per cent reduction in cooling-energy consumption. According to CEO Mayank Srivastava, this change not only reduces the carbon footprint but also delivers commercial benefits for tenants.

“Raising temperatures can lead to cost savings for customers because the amount they pay for power also comes down,” Srivastava explained. The initiative initially faced resistance from tenants but was eventually accepted through careful analysis of server tolerances and engagement with regulators.

Sustainable Solutions for the Future

Other companies are exploring even more advanced solutions. Nxera, for instance, opened a new data centre in Tuas that achieves a power usage effectiveness rating of 1.25, meaning only about 20 per cent of energy is lost to cooling, lighting, and other support systems. The facility uses smart thermal management, high-efficiency electrical systems, and direct-to-chip liquid cooling, along with solar power and rainwater harvesting to reduce its environmental impact.

ST Telemedia Global Data Centres takes a different approach, treating the constraints of the tropical climate as a design opportunity rather than a barrier. Its facilities are engineered with high-efficiency hybrid air and liquid-cooling systems, including closed-loop and recycled-water approaches that ease pressure on local water supplies.

Keppel, which operates 39 data centres with more than 800MW of gross power capacity, is exploring unconventional cooling methods. Its seawater-cooled floating data centre project in Singapore aims to serve as a proof of concept for markets where land, power, and water resources are limited.

A Broader Shift in Responsibility

While operators are making strides in efficiency, experts argue that the responsibility extends beyond them. Lock Kai Sang, head of the Singapore Institute of Technology’s Energy Efficiency Technology Centre, highlights the need for tenants to refresh ageing servers, adopt higher-efficiency equipment, and improve the efficiency of their software and AI models.

“For high-density AI workloads, raising room temperatures is a necessary but insufficient response,” he said. “The region will increasingly need targeted liquid cooling and tighter electrothermal designs.”

Governments also play a crucial role in setting clearer efficiency baselines for both new and existing data centres. Lock suggests that approvals for new capacity should be tied to credible efficiency and decarbonisation plans rather than granted solely based on scale.

Looking Ahead

As the demand for data centres continues to grow, the industry must move beyond simply “designed to be efficient” and focus on measured, operating-year performance. Lee Poh Seng emphasizes that governments should treat data centres as strategic infrastructure, considering their intersection with digital policy, electricity planning, water management, and climate policy.

“In the next phase of this industry, I do not think ‘designed to be efficient’ will be enough,” he said. Operators will increasingly need to show measurable performance in energy use, water consumption, and emissions.

Related posts