A wave of AI server farms is turning Malaysia into Southeast Asia’s digital epicentre, but their high energy use could derail the country’s climate goals, writes Genevieve Mallet, Dialogue Earth.
DATA CENTRES are springing up across Southeast Asia as China and the US vie for Artifical Intelligence (AI) supremacy. Tech giants – from ByteDance and Alibaba to Google and Microsoft – are setting their sights on Johor, Malaysia’s manufacturing hub just across the strait from Singapore. A rich agricultural state, Johor is now also home to sprawling data centre parks.
Since 2023, Malaysia has announced RM99 billion in data centre investments, much of it involving AI facilities, with a further RM149 billion in the near future. With cheap electricity, labour and ample land, Malaysia has proven to be a compelling alternative to its land-constrained neighbour and is set to become the fastest-growing data-centre hotspot in Southeast Asia.
(Feature image: Workers at a data centre under construction in Sedenak Tech Park in Johor | Image © Vincent Thian / Associated Press / Alamy)
Like many of its regional neighbours, Malaysia is positioning itself as an intermediary between the US and China amid their trade war.
However, this tightrope is growing more difficult to walk – a Chinese company was recently accused of renting Malaysian data centres via third parties, gaining indirect access to advanced, US-designed microchips that are restricted by US export controls.
Weeks after an investigation into that case was launched, the US announced that Malaysia will face 25% tariffs from August, though talks over a trade deal are ongoing between the two nations.
“We can expect a continual and delicate balancing act where Malaysia seeks to avoid taking definitive action that antagonises either the US or China,” says Ewan Lusty, a director at Flint Global, a firm offering policy and regulations advice to businesses.
Counting the carbon
The surge strengthens Malaysia’s role in the global data centre supply chain, but also raises environmental concerns. Shabrina Nadhila, Southeast Asia energy analyst at Ember, warns that the rise of energy-hungry server farms risks increasing emissions and “locking in fossil-fuel infrastructure that may become stranded”.
AI data centres require advanced computing to train and run large machine learning models, consuming significantly more energy than conventional data centres.
Energy-to-computing ratio
A key sustainability measure used by the industry is Power Usage Effectiveness (PUE). This ratio compares the energy used for facility operations, such as cooling, lighting and ventilation, against that for computing.
A perfect PUE score of 1.0 would mean that every watt powers computation alone. Hyperscale data centres, which host tens of thousands of servers, are more efficient and can achieve scores as low as 1.1.
But PUE only tells part of the story. “It’s an internal sustainability metric,” says Khoo Wei Yang, a research associate at the Khazanah Research Institute (KRI). He notes that PUE overlooks where the electricity comes from, how much is used overall, and the emissions tied to its generation.
Even the most technically efficient data centre can have a significant environmental footprint if it runs on fossil fuels. Khoo highlights that hyperscale data centres consume between 20-100 megawatts (MW) of power; “enough energy to power a small city”.
Up to half of the total energy consumption is for cooling the computer systems using vast air conditioning and cooling set-ups that use large amounts of water and energy.
In 2024, 81% of Malaysia’s electricity came from fossil fuels, much of it coal-fired. That reliance jars with corporate sustainability claims and is a challenge to the government’s National Energy Transition Roadmap, which targets 70% renewable capacity by 2050. Ember projects that data centre electricity demand will surge from 9 terawatt-hours (TWh) in 2024 to 68 TWh by 2030.
Tough for renewables
Malaysia’s fossil-fuel-heavy grid makes integration of renewables expensive and slow. It’s common for data centres to buy renewable energy certificates (RECs) to offset emissions, but “they could be using very dirty coal-fired power in Malaysia and then just buy cheaper RECs elsewhere”, says Khoo. Without standardised accounting, double-counting is a real risk, he adds.
Last year, Malaysia launched its Corporate Renewable Energy Supply Scheme (Cress), which allows large corporate consumers to procure renewable energy directly from independent power producers.
But uptake of the scheme has slowed because of added fees it has imposed on producers to access the network and help maintain it, meaning businesses are currently paying a premium for transitioning to renewable energy.
Can data centres drive the transition?
Some industry voices insist that data centres can be part of the solution. “Data centres actually play a critical role in accelerating the transition towards renewable energy by driving demand for clean power and investing in innovative energy solutions,” says Lusty.
He notes that Cress can help align energy-intensive sectors like data centres with national decarbonisation goals. Facility operators may also face pressure to meet environmental, social and governance (ESG) targets from global investors and clients.
Yet not all renewable efforts are created equal. Australian data infrastructure company AirTrunk recently announced a rooftop solar project at its JHB1 campus in Johor that will generate just 1 MW – less than 1% of the facility’s 150 MW load.
An oxymoron
This illustrates what Sara Loo, associate research officer at Singapore’s ISEAS-Yusof Ishak Institute, describes as the illusion of progress created by highly visible but marginal solar installations.
As Loo puts it, “this whole idea of a sustainable data centre is itself an oxymoron” because these facilities are inherently energy intensive, water intensive and carbon emitting. Despite growing pressure to go green, the core model remains environmentally costly, she notes.
then environment minister Nik Nazmi Nik Ahmad said earlier this year.
Malaysia had become “more selective” when approving new data centre projects because of water and energy constraints, he highlighted. But he said that tech companies are expected and “willing to pay a premium”, enabling them to “push boundaries” in renewable energy and water-recycling systems in ways that many other industries cannot afford.
Filling a gap
Unlike Singapore, which imposed a moratorium on new data centres from 2019 to 2022 to assess how it could manage its data centre growth sustainably, Malaysia is not likely to slow down.
“Johor is benefiting a lot from the moratorium that Singapore [had] in place,” Loo explains. But she cautions that Malaysia’s ongoing expansion into data centres, supported by an incentive-based approach focused on efficiency optimisation, may not sufficiently address the environmental and social impacts of energy and water consumption at the local and national level.
Loo notes that the long-term consequences of this growth remain uncertain.
Nadhila highlights Singapore’s successful policy approach, which includes grants for green building upgrades and research aimed at boosting data centre efficiency, as a strong regional example.
“Targeted incentives can accelerate adoption of clean technologies,” she says, but “establishing a clear sustainability framework from the outset is also critical; once data centres are operational, retrofitting becomes much harder”.
Related Stories
Banking on River Reserves to Protect Water Supply
Securing Water in a Harsher Climate
Coal Can Be Costly – Who’s Paying?
Pay us for our services to produce more stories like these. Note: This is not a donation; you are buying a service.
Comments are welcomed but shall be moderated. Do not use language that is foul, slanderous, violent or that may violate laws. Personal attacks will not be tolerated.