Malaysia aspires to provide the backbone of the global AI boom. Will the bet pay off?

Yahoo

18-04-2025

The jungles of Johor, the Malaysian state across the Johor Strait from Singapore, were first cleared in the 1840s by Chinese clans from Singapore seeking more space to grow black pepper. In the next century, under British rule, those pepper farms gave way to vast plantations of rubber and oil palm trees. On many of those same sites today, Johor is cultivating a new kind of cash crop: data centers meant to feed the world's voracious appetite for artificial intelligence.
Johor's data center boom, like the shift to growing pepper, is partly a function of scarcity in Singapore. The tiny city-state is Southeast Asia's digital hub. But it imports both water and power, and in 2019 imposed a moratorium on building data centers because the hulking facilities were guzzling water and consuming 7% of Singapore's electricity. Investors and operators of data centers flocked to neighboring Malaysia, where land is cheap, energy is abundant, and the government is eager to jump-start development of the nation's digital economy.
But Johor's rise as a data center powerhouse is also driven by the global scramble for computational power. Singapore rescinded its data center ban in January 2022, but the release of ChatGPT later that year triggered an explosion in global demand for AI infrastructure—and ignited a new investment frenzy in Malaysia. In 2023, Malaysia reaped more than $10 billion in investments for data centers, then tripled that in 2024, making the country the world's hottest destination for data center investments in both years, according to property consultancy Knight Frank.
Johor is the epicenter of that construction surge. For the state, and for Malaysia, the big question is whether this flood of capital and expertise will carry their broader economies to a new era of high-tech growth—or whether other challenges, like shifts in global demand and constraints in local resources, will turn their data centers from cash cows to liabilities.
Johor hosts more than 40 data centers that are either operational or under construction, according to advisory firm Baxtel, up from about a dozen in 2022. Many more are in the planning stages. Data center capacity, measured in how much power the facilities can provide, surged to over 1,500 megawatts last year, up from 10 megawatts three years earlier, according to data center market intelligence platform DC Byte.
If expansion continues at its current breakneck pace, Johor could overtake Northern Virginia as the world's largest data center corridor within the next five years.
'Johor is adding data center capacity at a speed and scale I've not seen ever anywhere else in the world,' says Rangu Salgame, CEO of Princeton Digital Group, a Singapore-based data center operator. Princeton Digital, whose backers include private equity giant Warburg Pincus, last year launched the first phase of a $1.5 billion, 150-megawatt data center campus in a massive tech park 40 miles inland from the Singapore-Johor border crossing—and plans to add a second, 200-megawatt campus at a business park a few miles up the road.
The parade of companies piling in with multibillion-dollar investment announcements in Johor also includes global tech leaders like Nvidia, Microsoft, Alphabet, and Oracle, plus data center operators such as California's Equinix, Japan's NTT Data, and China's GDS Holdings.
'Three years ago,' says Salgame, 'if you'd asked CEOs of the global tech giants about Johor, they'd have never heard of it, much less be able to find it on a map. Now, everyone is here.'
It's no accident that the advent of large language models (LLMs) in the U.S. and China has sparked a data center bonanza in faraway Malaysia. In the pre-ChatGPT era, for many of the services handled by data centers, there was an enormous advantage to operating from facilities physically close to end users. For functions like online gaming, stock trading, fraud detection, social media, or streaming videos, every millisecond counts. Companies providing such services pay huge penalties for 'latency'—sluggishness in the time it takes for data to travel between a user's device and the data center and back.
In contrast, training LLMs isn't interactive. Instead of sending requests and waiting for real-time responses, it involves running long, continuous computations on fixed datasets. The process can run for days or weeks without needing rapid back-and-forth communication. When latency isn't a concern, AI companies can instead prioritize efficiency—cheap and abundant power and land—and locate data centers thousands of miles from where the models are designed or meant to be used. That means Malaysia's AI data centers can compete not only with those in Singapore or other Southeast Asian neighbors, but also with similar facilities worldwide.
Malaysian Prime Minister Anwar Ibrahim has welcomed the data center boom and is rolling out strategic initiatives, including tax breaks and streamlined approval procedures, to position the nation as a global AI hub. A critical part of that push is the Green Lane Pathway, a 2023 initiative launched by Tenaga Nasional Berhad, Malaysia's primary electricity utility, that aims to reduce the time required to connect data centers to the power grid to 12 months, down from more than three years prior.
There are signs the data center boom is straining Malaysia's resources—for some of the same reasons the facilities were temporarily banned in Singapore. Malaysia, like Singapore, is one of the most water stressed countries in the world. Malaysia's National Water Services Commission has warned that the country could face widespread water shortages in the next five years owing to climate change and aging infrastructure—even without factoring in increased demand from data centers.
Power, too, is an issue. A medium-size data center might have a capacity of 40 to 50 megawatts, enough to consume as much electricity in a year as about 125,000 homes, depending on usage. Large hyperscale AI processing centers can require as much as 500 megawatts continuously, consuming more electricity annually than the roughly 250,000 households in Johor's largest city, Johor Bahru.
Malaysia's position on the equator means that its data centers also require far more energy to cool than facilities in northern countries with colder climates.
At a recent investor conference, Johor Bahru Mayor Mohd Noorazam Osman acknowledged concerns about water and power shortages. 'People are too hyped up about data centers nowadays,' he said. 'The issue in Johor is we do not have enough water and power. I believe that while promoting investments is important, it should not come at the expense of local and domestic needs of the people.'
The Malaysian government says it expects data centers operating in the country to pay a premium for water and power; early indications suggest tech companies and operators are willing to do so. Authorities last year rejected the applications of a handful of data center projects for failure to comply with efficiency and sustainability standards.
The rapid increase in power demand from data centers could prove a boon if it accelerates Malaysia's transition to renewable energy. In 2020, only about 4% of Malaysia's power came from renewable sources. That percentage is expected to rise above 30% this year, according to the Malaysian Investment Development Authority, and the government has vowed to increase the share of renewable energy in total generation capacity to 70% by 2050.
In Johor, authorities are thinking big. Princeton Digital's first data center campus is located in Sedenak Tech Park (STeP), a 700-acre digital hub owned by the property arm of JCorp, a state-owned conglomerate, on a site that was once part of a sprawling palm oil plantation. In addition to the Princeton hub, STeP includes a 300-megawatt hyperscale data center campus being built by Amsterdam's Yondr Group, and a third, under development by Japan's Mitsui & Co., that will include an on-site solar farm.
STeP, already Malaysia's largest data center complex, is about to get bigger. JCorp is developing a second phase, STeP 2, that will add another 640 acres to the park, and has plans for a 7,000-acre township that will include R&D facilities, residential areas, and culture and rec centers. JCorp also has engaged Zaha Hadid Architects to design a 500-acre innovation hub called Discovery City that will integrate digital technologies and sustainable living.
The proliferation of such projects is transforming Johor, Malaysia's southernmost state. Johor and Singapore are connected by two land crossings, Woodlands and Tuas Link, that are among the busiest and most congested border crossings in the world. The Singapore side is densely populated and carefully organized, with tolls and customs digitized. The Johor side is bustling and chaotic, with far more motorcycles, small cars, and buses.
"Johor is adding data center capacity at a speed and scale I've not seen ever anywhere else."
In January, Johor signed a 'special economic zone' agreement with Singapore to promote cross-border cooperation between the two economies. The motorbike-loving billionaire sultan of Johor, currently Malaysia's king under the country's rotating monarchy system, is championing the effort to bring his state and Singapore closer together. The agreement includes tax breaks, enables smoother cross-border trade, and makes it easier for skilled labor to move back and forth across the border.
It's unclear whether data centers will generate more and better jobs for Johor. Most data centers provide about 30 to 50 permanent jobs. Larger facilities might create as many as 200. But on their own, data centers seem unlikely to significantly boost overall wealth in Johor, where the GDP per capita is about $10,000 compared with nearly $85,000 in Singapore. Nor is it clear that Malaysia can use the development of data centers to attract other tech industries such as chip manufacturing.
The larger risk is a global data center bubble. The so-called DeepSeek Shock (China's breakthrough AI model that rattled Wall Street) could reduce the scale and demand for data centers everywhere if an overhaul of AI models to match DeepSeek's low-cost approach reduces demand for cutting-edge chips, expanded power plants, and large-scale data centers.
Salgame, for his part, says he is 'not the least bit worried' about flagging demand for computational power from the data centers Princeton Digital is building in Johor. Cheaper, more efficient AI models will only accelerate the world's use of AI—and the need for low-cost AI training centers in places like Johor. 'This is only the beginning,' he says.
This article appears in the April/May 2025: Asia issue of Fortune with the headline "Malaysia's data center power play."
This story was originally featured on Fortune.com