Published by BusinessToday, image by BusinessToday.
If it has not been said enough, data centres are extremely resource intensive. According to the International Energy Agency (2025), electricity consumption by data centres was estimated to be 1.5% of global electricity usage in 2024—amounting to 415 terawatt hours (TWh). However, experts suggest that electricity consumption in accelerated servers—primarily used for Artificial Intelligence (AI) purposes—could grow by 30% annually.
While 1.5% of global electricity consumption may seem insignificant, it is important to note that powering a single data centre can consume as much electricity as 100,000 households. Some data centres are even projected to consume up to 20 times that amount (The Guardian, 2025).
In Malaysia, the Deputy Prime Minister, and Minister of Energy Transition and Water Transformation, Fadillah Yusof, stated that our data centre electricity demand could reach 12.9 gigawatt (GW) by 2030 and rise to 20.9GW by 2040 (Malay Mail, 2025).
This is set to place a massive strain on our power grid, which is predominantly powered by coal and natural gas—two of the biggest contributors to global climate change. As EMIR Research previously pointed out, carbon dioxide emissions originating from data centres are expected to more than double between 2022 and 2030 (refer to “Unlocking Malaysia’s Nuclear Future: The BRICS Advantage”).
Electricity is not the only resource that data centres consume. Water is another vital input for certain data centres—used both for cooling and for generating electricity.
According to Mytton (2021), data centres consume water in two ways: indirectly through electricity generation and directly through cooling systems. In 2014, data centres in the United States (US) alone were responsible for the use of 626 billion litres of water. Further, Google’s average data centre consumed about 1.7 million litres of water per day, while data centres in Johor collectively consumed roughly 1.4 billion litres daily (The Edge, 2024).
Making matters worse, some of the largest data centre operators are building their facilities in regions already experiencing water scarcity (The Guardian, 2025a).
Water issues are localised, meaning efforts to manage data centre water usage in Malaysia will likely have little effect elsewhere in the world. However, they will have a significant impact on local Malaysian communities.
Despite receiving an average annual rainfall of 3,085.5 mm, Malaysia regularly experiences water supply disruptions (World Bank Group,n.d.). On top of that, reservoir levels often fall below safe thresholds, forcing operators to reduce water releases to conserve supply.
Given that the average Malaysian used 226 litres of water per day in 2023—and that Malaysian water demand is projected to grow by 103% by 2050—the growth of data centres in Malaysia presents a serious dilemma for policymakers and underscores the need for proactive measures to protect livelihood of the people (The Sun, 2025).
Reducing the water used for power generation is difficult, as we remain heavily reliant on coal and natural gas power plants. Without a major shift in our energy mix—possibly incorporating advanced technologies such as small modular reactors, which offer designs with significantly reduced water dependency—meaningful progress is unlikely (see EMIR Research’s earlier analysis, “Nuclear Power as an Option in Malaysia’s Energy Mix”).
Water usage within data centres, on the other hand, could be reduced through advancements in cooling methods.
Some of the most common cooling methods for data centres include air and water-cooled systems, each with its own advantages.
While air cooling uses significantly less water, its energy requirements are considerably higher than other methods (Haghshenas et al., 2023).
Liquid cooling is more effective at transferring heat and requires less electricity. However, as mentioned earlier, it consumes large volumes of water, making it potentially unsustainable for local communities.
This is why another cooling method—immersion cooling—is gaining traction not only for data centres, but also in other applications such as cooling car batteries in electric vehicles.
Immersion cooling involves submerging electronic components in dielectric liquids specifically designed to dissipate heat directly from the equipment, rather than relying on water. This method is characterised as highly efficient, low noise, environmentally friendly, and—when applied in two-phase immersion systems—especially beneficial for high-capacity data centres (Xu et al., 2023).
Experiments with immersion cooling have produced promising results. Hnayno et al. (2023) found that single-phase immersion cooling reduces electricity consumption by at least 20.7% compared to liquid cooling, and in certain data centres, even eliminates water usage entirely.
Although early studies suggested immersion cooling might not be cost-effective for data centres, later research has concluded that the technology’s low power requirements and minimal infrastructure needs can actually make it more economical than other cooling methods (Mukherjee et al., 2020; Pambudi et al., 2022).
Most dielectric coolants are hydrocarbon-based, typically derived from mineral oil. However, researchers have explored plant-based alternatives, including coconut oil and palm oils.
Pambudi et al. (2022) stated that while mineral oil outperformed virgin coconut oil in cooling efficiency, the later holds potential as dielectric liquids if its viscosity stability can be improved.
Akmal et al. (2023) on the other hand reported impressive results when comparing cooling effectiveness across various media—natural convention, air, 3M Novec 7000 (an engineered fluid used for cooling reactors in pharmaceutical and chemical processes), and palm oil. At an inlet velocity of 1 mm/s, palm oil emerged as the most effective, delivering the lowest maximum facet temperature over time and the highest heat transfer rate.
A recent study further reinforced palm oil’s potential as a heat transfer medium, revealing that base palm oil exhibits higher thermal conductivity than mineral oil. In fact, its heat transfer coefficient surpasses that of mineral oil mixed with 1% aluminium oxide (Hussain et al., 2025).
Of course, more research is needed to fully assess how vegetable oils can be optimised for use in heat transfer fluids. But the existing evidence points to real potential. Given that Malaysia is the second-largest palm oil producer—accounting for 24% of global output—there is a strong case for investing in further experimentation and development (USDA, n.d.).
Ultimately, while data centre expansion is beneficial for the economy, we must also prioritise environmental sustainability and the wellbeing of our people.
If alternative cooling methods exist that minimise environmental and social costs, we should actively encourage their exploration and adoption. Only then can technology truly serve the interests of society.
Chia Chu Hang is a Research Assistant at EMIR Research, an independent think tank focused on strategic policy recommendations based on rigorous research.
