Eric Yong has not enjoyed a proper weekend for a long time. A senior business development manager for engineering services company CNMy in Malaysia, Yong said he even had to field calls during the country’s Independence Day holiday as inquiries about investing in Johor flooded in from overseas.
“I feel like this is the beginning of the bright next decades for Johor and Malaysia as a whole. In the past we relied heavily on agriculture, but now we [have] boarded on the fast train to success,” Yong told Nikkei Asia as he stood in front of a cluster of data center construction sites in the southern Malaysian state.
“Here and there, the areas you are looking at have all been reserved and will be developed as new data centers,” Yong said, pointing toward the vast forest of palms surrounding the cluster. “Five years ago, there was no data center in Johor, but now everything has changed.”
“I am hosting so many different companies and tours from all kinds of sectors … data centers, construction, semiconductors, materials. … Everyone is looking to seize this paradigm shift opportunity,” Yong said.
The dramatic change comes against a background of global and regional upheaval: geopolitical tensions are spurring tech manufacturers to diversify away from China, while the surge in demand for generative AI since late 2022 has reshaped Asia’s data center landscape, challenging the longstanding dominance of Japan and Singapore.
The result of these two trends has been a flood of investment into Malaysia, India, Indonesia, and Thailand, from companies ranging from cloud service providers and data center operators to e-commerce platforms.
“If you look at Johor, in just the past three years, they have already reached a [data center] capacity of more than 900 megawatts, something that took Singapore, the original hub, 12–14 years to build,” Dedi Iskandar, Asia Pacific regional director at datacenterHawk, told Nikkei Asia.
Megawatts are widely used as a measure of a data center’s capacity because energy available determines how much equipment, including servers and cooling systems, a facility can support. Typically, any facility exceeding 30 MW is considered large-scale and 100 MW or more is a hyperscale data center. 1,000 MW equals 1 gigawatt (GW).
Industry executives and policymakers say that hosting powerful computing clusters can boost a region’s strategic importance. But the race to build artificial intelligence data centers is not for everyone.
Key tech hubs like Taiwan, Singapore, and South Korea have attracted far fewer new AI computing facilities in terms of the number and capacity of planned projects, due in part to land and energy constraints. Taiwan, for example, introduced strict rules on new data centers in northern Taiwan with a capacity above 5 MW in late 2023.
Jabez Tan, head of research at Structure Research, said hosting data centers is economically and strategically beneficial, even if they do not directly attract large amounts of talent or create many permanent jobs.
“Once your country gets momentum as a data center hub, the railway of AI, it will start to create its own flywheel effect, where you’ll start to see more development happening in the country,” Tan told Nikkei Asia. “It will generate a certain level of velocity to attract connectivity infrastructure into your country, including submarine connectivity cables. … That’s how Singapore became such a strategic hub.”
Malaysia has gone from a marginal player to one of the region’s biggest magnets for digital investment in a few short years, attracting a total of at least MYR 210.4 billion (USD 51.4 billion) in 2023 and 2024, government data shows. Johor has claimed the bulk of that surge, though Kuala Lumpur and Cyberjaya are also home to emerging hubs.
A major factor in its rise as a data center hub is its proximity to Singapore, according to Tan, “because you can still connect back to Singapore via Johor terrestrially.”
It is just a 40-minute drive from Singapore’s bustling city center to the border with Johor. Rush hour traffic can increase travel time significantly, but a cross-border train set to start operation at the end of 2026 is expected to alleviate that congestion. The Johor-Singapore Special Economic Zone set up in January 2025 is expected to further strengthen business ties between the state and its neighbor by providing incentives ranging from faster immigration clearance to simplified custom regulations.
“We evaluated both Penang and Johor for our first factory outside of Greater China and I told my boss we should choose Johor,” said Steven Tan, an executive with Taiwan’s FIC Global Inc. “I strongly believe [in] the potential here brought by the proximity to Singapore, the benefits of the special economic zone and the rising data center clusters.”
FIC chose a new industrial park in Johor just 30 minutes from the border to build two manufacturing facilities for optical modules components. The first one became operational in September 2025, and the company is now focused on ramping up production there in hopes of playing an important role in the global data center supply chain, according to Tan.
RELATED ARTICLE
Peter Huang, global president of thermal management and data center at Castrol, also has his eye on Johor.
“In terms of AI data centers in Asia, the biggest ‘golden triangle’ we are looking at would be Johor, Singapore, and Batam,” Huang told Nikkei Asia. Relatively lower construction costs, especially in Malaysia, is one of the reasons new projects are coming into the region, he said.
Driving northward from the border, tech parks are sprouting up among Malaysia’s palm forests and mountains. Large and midsize cloud service providers and data center operators from China, such as ByteDance, Tencent, and Global Data Solutions, and the US, including Microsoft and Oracle, all have a presence here.
Sedenak Tech Park is the largest in Johor, covering 745 acres (around 300 hectares) in the Kulai district. Once home to rows of textile factories, the area is now filled with the hum of bulldozers and excavators. Its transformation into a data center hub is supported by critical infrastructure, including a high-voltage power distribution facility owned by Malaysian state electric company Tenaga Nasional Berhad and a reclaimed water plant within the park. Only a single textile plant remains in the entire complex.
YTL Green Data Center Park, next to Hutan Lipur Gunung Pulai, a forest famous for hiking and camping, is owned by Malaysian utility provider YTL Group. YTL is building the data center with Nvidia’s AI servers and aims to play a key role in supporting Malaysia’s aims for sovereign AI, or the ability of a country to independently develop, deploy and govern AI systems within its borders.
“The age of AI has given birth to new types of business opportunities, including neo cloud service providers and sovereign AI, which we are seeing YTL as a prime example for the Malaysian market,” said Emily Hong, CEO of Wiwynn, an Nvidia and AWS supplier that has built a server manufacturing facility in Johor. Wiwynn is also a strategic partner of YTL, helping it build AI data centers.
Malaysia’s data center boom has not only reshaped investment flows, it has also created jobs that pay well above the national average, according to Bank Negara Malaysia. Running and maintaining the facilities require specialists in electrical engineering and telecommunications, as well as network specialists, cloud architects and data scientists whose salaries are notably higher. Entry-level roles typically pay MYR 3,000–7,000 (USD 733–1,710) a month, compared with a national median of about MYR 2,793 (USD 682), while more experienced positions can reach MYR 10,000–30,000 (USD 2,443–7,329), according to the central bank’s estimate.
But the massive data center buildout also has its darker side. For one, job creation is far lower than in manufacturing, largely because of automation and the remote locations of many facilities. Data centers also place heavy demands on resources, with electricity and water already becoming hurdles for Johor.
An investment review committee under the Johor state government recently implemented a measure that rejects new applications for data centers if they call for liquid cooling systems due to water supply constraints.
Johor state executive councillor Lee Ting Han told Nikkei Asia that Johor has 15 operational data centers, 11 under construction and 25 approved projects, for a combined capacity of 5.3 GW.
“We are simultaneously building three water supply facilities to cater to the demand, and the first phase will come [online] by June 2027, so we told new applicants to change their cooling solutions or we will reject their applications,” Lee said. “We have enough power and water, but for water we need more supply system infrastructure and it takes time.”
When choosing a location to build AI data centers, the decision is often less about ambitions or market potential and more about “whether power and land are available,” Shun Endo, head of data center capital markets for Asia Pacific at CBRE, told Nikkei Asia.
In Japan, for example, it is no longer possible to secure both land and sufficient electricity near central Tokyo or Osaka, Endo said. “It’s very difficult for AI data centers to exist in these areas simply because power and land are scarce.” In Australia, by contrast, land is plentiful, but energy availability can still pose challenges, he added.
According to Endo, Johor is gaining traction mainly because of spillover from Singapore, where both land and energy are constrained.
Geopolitics and export controls are another risk. Iskandar of datacenterHawk told Nikkei Asia that access to AI chips remains a major unknown as competition intensifies to become the region’s next data-center hub.
“If a new investment is coming into Malaysia, Indonesia, or Thailand and there are no certainties whether they are able to bring in the AI chip, there will be a big risk that once they build the facilities, there will be no takers,” Iskandar said.
Malaysia launched a national AI office at the end of 2024 and set a goal of becoming an “AI nation” by 2030. It is striving for a neutral stance between the US and China, welcoming investments from tech companies from both sides.
But risks exist. Malaysia, like Singapore, has more than once come under scrutiny from US policymakers concerned that its data centers could be used by Chinese tech giants to avoid Washington’s export controls on chips and other advanced technologies.
In response, the Anwar Ibrahim government in July imposed new export controls on high-performance AI chips of US origin to prevent attempts to “circumvent export controls or engage in illicit trade activities by any individual or company.”
Lee said the Malaysian government hopes to maintain a balance between the world’s two superpowers and that mindset applies when approving new data center projects in Johor.
“It is important to keep neutrality while we pay a very close attention to the export control measures,” he said. “The access of Nvidia’s AI chips would have a profound effect on the sustainability of the AI data center trend in Malaysia.”
Another source of concern is oversupply. Unlike cloud giants such as Google, AWS, and Microsoft, which build AI data centers to support their own services, many carrier-neutral operators are taking a gamble. Well-known operators Bridge, Equinix, Vantage, Keppel, Princeton, GDS, and the Kwok Group have all invested heavily in AI-ready facilities, and there are many new GPU cloud service providers tapping into the field. But much of their capacity still needs to find enough tenants, meaning they are likely building ahead of demand, according to multiple analysts and industry executives.
There have already been some casualties. Real estate companies UEM Sunrise and Logos have scrapped plans for a 350 MW data center in Johor, citing a change of business strategy and direction, just a year after the announcement.
Peter Wu, CEO of supercomputer builder Taiwan AI Cloud, a subsidiary of Asustek Computer, said Malaysia has relatively cheap electricity prices and abundant land but not much domestic demand for GPUs.
“The AI data center demand in Malaysia is mainly from the perspective of traditional CSP [cloud service provider] approach, which is to find a cost-effective location,” said Wu, who helped the Taiwanese government build the Taiwania supercomputer. He added that he has not yet seen a lot of sovereign AI cases by the government and enterprises in Malaysia.
“Sovereign AI may be healthier for AI development. … It is a government investment to accelerate local innovations or for startups or enterprises to use,” he said, adding that there is no return on investment issues for sovereign AI projects. Commercial endeavors, by contrast, “sell GPU resources to rich customers, and in the Asian market, there are not many rich customers or rich startups like in the US.”
Iskandar of datacenterHawk said Malaysia is currently the frontrunner in attracting data center investment but needs to create “more local demand” to make the overall environment sustainable. Other Asian nations such as Thailand, Vietnam, Indonesia, and India, with larger land availability, lower costs and bigger populations, are positioning themselves to appeal to the same pool of investors.
Some of the buildup in Johor, he said, “is not based on strong fundamentals of local demand so … there is a fear of oversupply.” Eventually, he believes, capacity will be distributed based on market demand.
“I don’t think there will be a single country to become an AI hub. I think everybody is just chasing which locations have the proper infrastructure to handle AI specification [computing],” Iskandar said.
Castrol’s Huang said other contenders are emerging, such as Australia, which carries less geopolitical risk and does not have to deal with restrictions on access to advanced AI chips, while also benefiting from solid demand from sectors like corporate IT and mining.
Others take a different view. “We are still in an undersupply situation,” Robert Cheng, head of Asia tech hardware research with BofA Global Research, told Nikkei Asia. “Every sentence, every video AI produces relies on computing power, and there is room for AI to improve. … The demand is huge.”
For Yong, the construction management consultant and longtime Johor resident, Malaysia’s AI data center boom is a chance for the nation to elevate both its economy and its global stature.
“There are certainly challenges, such as a shortage of construction workers … but this is a once-in-a-lifetime opportunity we have to seize,” Yong said.
This article first appeared on Nikkei Asia. It has been republished here as part of 36Kr’s ongoing partnership with Nikkei.