Artificial intelligence is evolving rapidly, and so is the demand for the infrastructure behind it. From the explosive growth of generative models to the rise of AI-driven data analytics, companies are scaling up data centers at an unprecedented pace. But for the sustainably-conscious, this rapid growth sparks a real “down the rabbit hole” question, “what is AI’s environmental footprint?”
The focus on AI’s sustainability has largely centered on energy use, and rightly so. Powering massive workloads takes a toll on our grids. However, another important (yet overlooked) piece of the sustainability puzzle is the environmental impact of maintaining these systems, particularly when it comes to cooling.
As rack densities and high-performance chip use increase to accommodate the growing AI demand, effective cooling becomes essential to maintain uptime and performance. According to the U.S. Department of Energy, cooling alone can account for up to 40% of a data center’s total electricity consumption. To meet both operational and environmental goals, data center operators need to reevaluate how they manage the heat.
Data Center Cooling Today
A major shift we’re seeing is the move from reliance on air-cooling systems to liquid cooling solutions. Traditional air cooling used to do the job. However, it is no longer sufficient for today’s AI demands. Now, liquid cooling systems are filling in the gaps and proving to be far more efficient.
Techniques like chilled water loops, direct-to-chip (cold plate) cooling, and immersion cooling are gaining popularity. These liquid cooling approaches handle heat more effectively, reduce the size of chiller equipment, and support higher-density computing.
“Liquid-cooled data centers enable up to 50% more compute with higher core densities with the same or even less IT power than air-cooled data centers (480,000 cores versus 320,000 respectively),” said David Gyulnazaryan, director and principal consultant at Impleon, a leader in data center decarbonization and heat reuse. “This shift not only elevates operating temperatures of facility water from 16C to as high as 44C, but also significantly reduces the power and space needed for cooling, leading to a reduced reliance on refrigerants.”
But here’s something worth spotlighting—while switching to liquid cooling helps with energy efficiency, the fluid itself also matters.
The Limitations of Fossil-Fuel Cooling Fluids
While data center leaders are familiar with optimizing server efficiency and power usage effectiveness, they may not be giving enough attention to the choice of cooling fluid and its significant environmental impact. Traditionally, petroleum-based glycols, such as propylene glycol (PG) and ethylene glycol (EG), have been used in thermal management systems, and although these fossil fuel-based fluids are reliable, they come at a heavy environmental cost.
Their production is energy-intensive, contributes to greenhouse gas emissions, and in the case of EG, often requires special handling due to its toxicity. As a result, even if a data center adopts a more efficient cooling method, sticking with fossil-based fluids undercuts some of the sustainability gains.
The Sustainable Solution
Fortunately, there has been growth in sustainable alternatives to traditional cooling fluids. Bio-based heat transfer fluids sourced from renewable materials instead of petroleum are opening the door for improved sustainability without compromising performance.
Like petroleum-based glycols, bio-based heat transfer fluids can be used in a range of liquid cooling configurations, including direct-to-chip, cold plate systems, and facility primary loops supporting immersion setups. Additionally, some fluids like Susterra propanediol, a 100% bio-renewable glycol sourced from industrial dent corn, can be used as a glycol/water mixture to cool immersion tanks externally, even though they are not immersion fluids themselves. Their chemical stability at high temperatures also helps minimize thermal degradation byproducts like glycolates and slows nitrate depletion from oxidation, which supports longer fluid life and a more consistent system performance.
In combination with some system designs, these fluids can reduce pump power, lower flow rates, or even eliminate the need for chillers, saving both energy and water. By replacing fossil-derived glycols with renewable alternatives like Susterra propanediol, data centers can also report lower greenhouse gas emissions in their sustainability metrics and ESG disclosures.
Every Choice Matters
It’s easy to overlook fluid choice as a small detail in the bigger picture of AI infrastructure. But whether you use air cooling, liquid cooling or a hybrid of both, as the industry pushes toward more sustainable practices, every component adds up and impacts your sustainable goals. Choosing bio-based thermal fluids is one actionable way to cut environmental impact, without sacrificing reliability or efficiency.
AI isn’t slowing down, and neither is its influence. As we scale to meet demand, we have a responsibility to build smarter. Whether it’s in hardware design or the fluids that keep systems running, thoughtful decisions will shape the future of a greener, more resilient digital world.
—Jessica Gallagher is Susterra Marketing Manager, Americas, at Primient Covation.
