Why Southeast Asia’s AI Data Centres Need a Better Way to Measure Sustainability

While the environmental toll of artificial intelligence (AI) is increasingly acknowledged, the way we measure the sustainability of the infrastructure powering it may be outdated. In some cases, it may even be misleading.

Traditionally, Power Usage Effectiveness (PUE) has been the industry standard for evaluating data centre efficiency. It measures the ratio of total facility energy to the energy consumed by computing equipment. In Singapore, where the government has set a PUE cap of 1.3 for new data centres, the metric is widely used to set regulatory targets.

But PUE alone paints an incomplete picture. It excludes factors like water consumption, electronic waste, and the carbon footprint of hardware production — all of which are becoming more pressing as data centre activity intensifies across Southeast Asia, with companies investing billions into the region. Proper ways of measuring data center inefficiencies are especially pertinent as the environmental impacts of energy and water use are often felt at the local level.

One such metric gaining traction is Water Usage Effectiveness (WUE), which captures the volume of water used per kilowatt-hour of IT energy consumption. French cloud computing company OVHcloud is one company in the region demonstrating how WUE can work in tandem with PUE to drive more comprehensive sustainability outcomes.

At its Singapore data center — its most sustainable facility in Asia-Pacific — OVHcloud reports a PUE of 1.26, meeting national requirements, and a WUE of just 0.37 litres per kWh, five times lower than the industry average.

This performance is driven by its proprietary closed-loop water cooling system, developed over two decades. Unlike traditional room-wide air-conditioning, the system cools servers at the chip level, with a water-cooled thermoblock sitting on the heat generating chip, circulating water in a closed-loop without evaporation.

The model has proven scalable, with OVHcloud deploying similar systems in its 5 other APAC data centers. Its Singapore branch remains the most efficient, despite not being its most recently set up center, providing an example of how infrastructural support, made available by government regulations, can aid technology. 

Beyond WUE and PUE, experts have also called for greater industry attention on Carbon Usage Effectiveness (CUE), which measures total greenhouse gas emissions per unit of IT energy. This includes Scope 1 to 3 emissions, making it a more complex and holistic indicator of sustainability. The ideal CUE is zero, signalling carbon neutrality. 

OVHcloud is notable as it manufactures its own racks and servers, allowing more clarity in its carbon reporting and culminating in a CUE of 0.16.

“With APAC’s data centre market projected to hit US$71.7B by 2032, the future belongs to infrastructure built with full-lifecycle responsibility in mind,” said Terry Maiolo, Vice President & General Manager, APAC at OVHcloud. 

“At OVHcloud, we don’t just run data centers, we design, build, operate, and recycle them. This circular model drastically reduces emissions and resource use. The next frontier in cloud isn’t just faster or smarter, it’s sustainable at its core. And Southeast Asia has a chance to lead that transformation.”

As Southeast Asia’s AI infrastructure continues to expand, there is a growing imperative to evolve the way sustainability is defined and measured, allowing economic opportunity while diminishing harm. Utilising PUE, WUE, and CUE provides a more holistic assessment of a data center’s impact, and provides a greater opportunity — one that enables smarter, greener cloud computing.