Sustainable Data Centre Evolution Strategies for the Physical AI Era

By 2026, the Asia–Pacific region has firmly established itself as the world’s most dynamic hub for digital transformation. In particular, Korea has gained a strategic edge in the global AI infrastructure race. Following the APEC summit last October, the country saw tangible progress in securing GPU supply commitments and new data centre investments from global leaders such as NVIDIA and Amazon Web Services (AWS). Korea is now accelerating not only the expansion of its digital infrastructure footprint but also innovation in operational efficiency, resilience, and sustainability.

The AI industry is rapidly evolving beyond generative AI toward “physical AI” — the application of AI across manufacturing, logistics, robotics, and other real world sectors. Demand is expected to surge across the entire industrial landscape. At CES 2026, AMD CEO Lisa Su highlighted the dawn of the “yottascale computing era,” while NVIDIA CEO Jensen Huang underscored that the coming years will be a defining turning point in the global AI infrastructure race.

At the core of this AI era lies the data centre. Yet reducing data centres to “buildings that house servers” misses their true nature. AI ready data centres function more like integrated power and thermodynamics systems built to support high density GPU clusters. As rack level power consumption increases, facilities evolve into sophisticated plants that integrate server rooms with substations, power distribution, UPS systems, generators, and advanced cooling technologies.

According to IDC Korea, domestic data centre power demand is projected to grow at an annual average of 11% through 2028, surpassing 6 gigawatts. The challenge is that this rising demand is more than an energy use issue: higher electricity consumption directly correlates with increased carbon emissions. As a result, data centres are becoming both the foundation of AI adoption and a critical proving ground for sustainability.

However, the outlook for data centres is not defined by concern alone. Operators are transforming environmental responsibility into a source of competitive advantage through clear, measurable roadmaps. STT GDC, for example, is progressing toward its 2030 carbon neutrality goal, already achieving more than 78% renewable energy usage across its global operations. In Singapore, the company has begun transitioning a portion of its emergency generator fuel from conventional diesel to hydrotreated vegetable oil (HVO), a renewable alternative that reduces lifecycle carbon emissions by up to 90%. Through AI powered operational control pilots, it also expects to achieve up to 30% energy savings—demonstrating that sustainability can be realized through tangible, data driven operational innovation rather than aspiration alone.

This momentum extends beyond data centre operators. Companies across the AI ecosystem are moving in the same direction. Semiconductor manufacturers are improving architectural efficiency, cloud providers are accelerating innovation in liquid cooling, and governments are embedding carbon requirements into procurement and policy frameworks. Together, these forces are creating a positive cycle in which environmental concerns become catalysts for sustainable innovation.

At CES 2026, NVIDIA introduced its next generation “Vera Rubin NVL72” architecture, delivering twice the energy efficiency of the previous generation and signaling a shift toward truly high performance, low power computing. Google and Microsoft have also adopted liquid cooling and microfluidics based approaches to reduce power consumption. Liquid cooling, capable of cutting cooling related energy use by more than 50% compared to traditional air cooling, is emerging as a key technology for improving energy efficiency and reducing carbon emissions.

Despite the optimism surrounding AI’s economic and societal benefits, concerns remain around the power and carbon footprint of the data centres that enable it. Yet as high efficiency compute, liquid cooling, and AI driven facility operations become more widespread, the industry can establish an energy efficient and resilient ecosystem that simultaneously optimizes power, heat, carbon, and operations.

Ultimately, the critical task is to embed efficiency and accountability into both infrastructure design parameters and operational KPIs—and to execute against them rigorously. If achieved, data centres will not only serve as the heart of the AI economy but will also transform environmental concerns into drivers of sustainable innovation, forming the resilient foundation for the digital future ahead.

This article was published on ZDNET Korea on March 3, 2026.

###