Race to green data centers: Three low-carbon trends as AI reaches GW scale

As AI drives up energy demand, Taiwan must focus on low-carbon energy use and technological innovation in data centers. (Image: iStock)

Global digitalization is driving a surge in demand for AI data centers, which have become major sources of energy consumption and carbon emissions. With the global energy transition accelerating and companies under mounting pressure to meet net-zero goals, decarbonizing data centers has become an urgent task. This series, “Race to green data centers,” explores how businesses, technologies, and markets are shaping solutions and charting the path toward sustainable digital infrastructure. 

As AI advances rapidly, power demand from data centers is rising sharply. Combined with the growing pressure on businesses to meet net-zero targets, existing energy infrastructure is struggling to keep up. To adapt, AI-driven data centers are expected to shift from traditional centralized energy generation to distributed energy systems, prompting changes in equipment and system design. In many ways, AI is sparking in an energy revolution in the data center industry. 

Trend 1: Rising AI power demand drives upgrades in energy equipment and technology 

As AI data centers scale up, GPU power consumption is rising sharply. According to DIGITIMES analyst Yu Pei-ju (余佩儒), a single AI chip is expected to consume more than 2 kW, pushing overall demand from the MW-scale into the GW level. Companies including OpenAI and Meta are already making preparations, with GW-scale AI data centers projected to come online as early as next year. 

As a key player in the AI supply chain, Taiwan must closely track the shift toward low-carbon energy and advances in data center technologies as electricity demand continues to rise.

Cooling has become a central focus in data center design. As power usage efficiency (PUE) standards grow increasingly stringent, the traditional reliance on air cooling is giving way to liquid cooling solutions. Power supply systems are also evolving, shifting from server-level units embedded in racks to independent power cabinets that can be paired with backup battery units (BBUs) for more stable electricity supply. 

To improve the efficiency of power conversion, grid-level innovations such as high-voltage direct current (800V DC) and solid-state transformer (SSTs) are expected to see wider adoption within the next two to three years, according to Yu.  

The future of energy is expected to move toward carbon-free and more diverse sources of green power. Yu noted that data centers require a stable energy supply, which makes baseload low-carbon energy increasingly important alongside wind and solar. One emerging model is co-location, in which data centers are built alongside power plants, allowing companies to source electricity directly without relying on gird distribution. 

Global data center electricity consumption, by equipment. (Source: IEA)