As AI reshapes industries at breakneck speed, the infrastructure powering this revolution, data centres are scaling up at an unprecedented rate. But while AI drives explosive growth in computing capacity and capital investment, it also brings with it urgent sustainability challenges that operators can no longer ignore.

According to GlobalData’s latest Strategic Intelligence report, titled “Data Centres,” the sector is under pressure to balance rapid expansion with responsible energy use, particularly as AI workloads demand massive computational power. These workloads are largely concentrated in large-scale, high-performance facilities that rely on specialised chips, advanced memory systems, and sophisticated cooling technologies.

“They are competing to build large data centres and fill them with specialised chips to stay ahead in the race for AI dominance,” said Martina Raveni, Analyst in the Strategic Intelligence team at GlobalData.

Major tech firms are leading the charge. Microsoft, Alphabet, Amazon, and Meta reported a combined capital expenditure of $245 billion in 2024, with projections indicating this figure could exceed $360 billion in 2025, driven largely by AI development.

Data centres have become foundational to the digital economy, supporting everything from enterprise software to cloud services. However, their growing energy footprint is raising alarms. The International Energy Agency (IEA) estimates that electricity consumption from data centres will more than double from 415 terawatt-hours (TWh) in 2024 to around 945 TWh by 2030.

Cooling systems and water use add further strain to environmental resources.

“Tech giants and data centre operators are increasingly shifting to low-carbon energy solutions to meet growing power demands within their data centres,” Raveni noted. “Companies that develop and deploy innovative cooling technologies, particularly liquid cooling, will likely see increased demand.”

As AI accelerates, sustainability innovation will be critical to ensuring the data centre boom remains both scalable and responsible.