- Blog Categories: Sustainability
As artificial intelligence places new demands on digital infrastructure, Steven Santini, secure power vice-president, Schneider Electric sub-Saharan Africa, explains why liquid cooling is becoming central to the future of resilient, efficient and sustainable AI data centres
Why AI is increasing data centre energy demand
Think about the last time you typed a question into an AI tool. That single query used roughly ten times more electricity than a standard internet search. Now multiply that by the millions of queries happening every second around the world, and you start to get a sense of the pressure building inside data centres right now.
Data centres currently consume roughly 1.5 to 2% of global electricity, and the International Energy Agency projects that demand will more than double by 2030. By that point, data centres are projected to account for approximately 3% of total global electricity consumption, growing four times faster than electricity demand from every other sector combined. The heat that comes with all of that computing power has to go somewhere. And that is where the real problem begins.
AI applications rely heavily on GPUs and accelerated computing systems that generate concentrated heat loads far greater than standard enterprise servers. For years, blowing cold air through server rows was enough. But AI workloads are a different beast entirely. Today, rack power densities already range from 40 kW to well over 100 kW. A fully populated NVIDIA GPU rack draws around 132 kW. The next generation of hardware, expected within the year, is projected to hit 240 kW per rack, and the industry is already looking ahead to a future where 1 MW per rack becomes reality. At those levels, air cooling is not just inefficient. It is unworkable.
The role of liquid cooling in AI infrastructure
Direct-to-chip liquid cooling removes heat at the exact point where it is generated, before it has a chance to spread. By capturing it directly at the chip level, liquid cooling is up to 3,000 times more efficient at removing heat than air. It can reduce a facility's overall energy use by 30 to 60 percent, and because it operates as a closed-loop system, it can eliminate water consumption from the cooling process entirely, something adiabatic air cooling simply cannot do.
Cooling already accounts for one of the largest shares of energy use in any data centre, second only to the IT load itself. Making it more efficient is one of the most direct ways to lower emissions, reduce costs, and meet sustainability commitments. For operators facing rising energy prices and growing environmental accountability, those gains are not marginal. They are structural.
Key benefits of liquid cooling in AI data centres:
- Higher energy efficiency
- Reduced cooling costs
- Lower emissions
- Better performance for high-density workloads
But switching to liquid-cooled infrastructure is not just a technical upgrade. It requires IT and facilities teams to plan together from the very beginning, aligning hardware decisions with facility design, power capacity, and long-term sustainability goals. The most expensive mistake an operator can make right now is letting AI hardware arrive before the infrastructure is ready to support it. Hybrid setups that combine air and liquid cooling, or chillers designed for higher operating temperatures, give facilities the flexibility to adapt across hardware generations without costly retrofits. Early collaboration with vendors, cooling specialists, and system integrators is what turns good intentions into working facilities.
Future-ready AI infrastructure will not be measured only by computing power. It will also be judged by efficiency, resilience and environmental responsibility.
Sustainability challenges for data centres
That urgency is being matched at a global level. According to the World Green Building Council, nine leading organisations from the built environment and sustainable finance sectors have launched the Greening AI Data Centres Coalition (GADCC) on 22 April 2026. The coalition is aimed at creating credible global benchmarks for greener data centre development, helping investors, operators, communities and policymakers distinguish measurable progress from broad marketing claims.
The IEA projects global data centre electricity demand will more than double by 2030, reaching approximately 945 TWh per year. Goldman Sachs estimates that around 60% of new data centre demand could rely on fossil fuels, adding approximately 220 million tonnes of CO2 globally. The IEA also warns that 20% of planned data centre projects are already at risk of delays due to grid constraints. Water is equally at risk, with some large data centres consuming up to 5 million gallons per day, equivalent to the daily needs of a small town.
The GADCC will focus its initial work on two priorities: developing an internationally aligned framework of environmental and social performance standards covering energy, carbon, water, waste, biodiversity and community impact; and supporting credible green finance instruments, including green bonds and sustainability-linked loans, for data centre investment that meets its standards.
The founding members are:
Building Research Establishment (BRE), Climate Bonds Initiative, German Sustainable Building Council (DGNB), Global Real Estate Sustainability Benchmark (GRESB), Green Building Council of Australia (GBCA), Green Building Council South Africa (GBCSA), Indian Green Building Council (IGBC), U.S. Green Building Council (USGBC), and World Green Building Council (WorldGBC).
The AI data centres being built today will be held to a higher standard than anything that came before them. The operators who recognise that now, and plan around it from the start, will be the ones best placed to lead what comes next.
