AI infrastructure can lead South-east Asia’s next growth story: UOB

[SINGAPORE] As artificial intelligence reshapes the business landscape worldwide, its underlying infrastructure is drawing attention as a crucial enabler of AI-driven growth.

For South-east Asia, the AI wave goes further than enabling businesses to grow revenues through adoption of the technology. Rather, the region can tap growing demand for infrastructure within the AI value chain to generate returns.

“The AI story is not just a technology story,” said Edmund Leong, head of group corporate banking and sector solutions group at UOB . “It is also an economic growth story.”

He made these remarks during a panel showcase last Friday (Jul 31) at the Asean Conference 2026, organised by the Singapore Business Federation.

He said that businesses have increasingly adopted AI models to improve operations and workflows across multiple sectors, from healthcare and financial services to manufacturing.

Deploying those models requires immense computing power, Leong told The Business Times in an interview during the conference.

“That compute comes from AI servers that train and run these models, which in turn drives investment throughout the ecosystem.”

Those components ultimately feed into data centres, which have risen to the forefront of the AI infrastructure ecosystem. Such data centres have accelerated in growth throughout South-east Asia.

Much of the region’s recent data centre growth has shifted towards Malaysia; Thailand, Indonesia and Vietnam are also building capacity, Leong noted.

SEE ALSO

South-east Asia’s data centres need to be planned like public infrastructure, not buildings

But the ecosystem will have to support these data centres through sufficient power generation, cooling capability and sourcing land for construction – all of which require funding.

Therefore, financing that infrastructure may prove just as important as developing the technology itself, he said.

Financing the build-out across Asean

But Leong flagged that capital for AI infrastructure projects does not just come from lenders. For financial institutions such as UOB, that means playing a role that extends beyond lending.

Describing the bank’s role as a “superconnector” of capital, he explained that financial institutions help projects mobilise capital through loans, or tapping capital markets through bonds or equity.

As AI infrastructure expands across multiple markets, financial institutions with regional footprints – including UOB – are becoming increasingly important in facilitating cross-border investment.

“Many of these data centres are owned by operators who manage multiple locations,” Leong noted, adding that this comes with large amounts of capital flows across borders.

Not every project within the AI infrastructure ecosystem is a bankable one that matches an institution’s appetite, he said.

“We need to work with clients who genuinely understand the business, have the technical expertise to execute the project and, importantly, have committed shareholders.

“AI infrastructure is extremely capital-intensive, so long-term commitment from investors is critical.”

Beyond financing, Leong said that many parties have to contribute to see a project through from start to finish.

He noted that projects often have to involve local regulators, utility services to ensure sufficient power and water availability, as well as telecommunications companies, to ensure supportive data networks.

Banks such as UOB therefore become “enablers” that connect these various parties together to enable a project to succeed, he said.

Within the region, many of these projects are spearheaded by large hyperscalers with the ability to serve clients across a variety of locations or needs, he added.

These hyperscalers, or cloud service providers, are also investing significantly to support rapidly developing AI systems – particularly in agentic AI.

“What’s most significant is the expanding scope of what agents can handle,” explained John Kain, director of financial services and market development at Amazon Web Services.

“We’ve moved from agents answering questions to agents owning entire workflows end to end.”

He told BT that these systems require infrastructure designed to support such operations – including data centres powered by renewable energy, or custom chips that are purpose-built for greater efficiency.

Translating AI into enterprise growth

Building AI infrastructure alone, however, does not guarantee stronger business performance. Both Leong and Kain said the next challenge for enterprises is translating AI investments into revenue growth.

“What we don’t want is people to just invest and be speculative about AI without knowing what the end goal is,” Leong said.

Kain, who spoke on the panel with Leong, told BT that companies will need clear implementation strategies to ensure AI is used effectively to generate productivity gains.

He said the companies scaling AI the most successfully are those which focus first on business outcomes before aligning these goals across the organisation.

“That focus helps them ensure they are investing not just in technology, but in the people and processes that make it work,” said Kain.

He added that successful organisations also build governance into AI deployment from the outset, allowing employees to experiment safely while ensuring new applications can be deployed at scale.

“Governance promotes innovation rather than constraining it. They’re already compliant when regulatory frameworks crystallise.”

Regional bottlenecks

Leong said that South-east Asia must overcome two bottlenecks to allow AI infrastructure deployment to flourish.

One of these is power. The region, he noted, will need to invest more in renewable energy and grid infrastructure to enable its data centre growth to accelerate.

He estimated that around US$150 billion could be invested into regional energy infrastructure over the next five years, across renewables and other energy transition projects.

“Asean is not short of resources,” he said. “The big vision of Asean is to essentially optimise where power goes.”

The other constraint is semiconductor supply.

“There is currently a shortage of semiconductor and memory chips,” he said, noting that strong demand continues to outpace manufacturing capacity.

But he expects technological improvements and expansion in supply to gradually lower costs for the infrastructure build-out while also improving efficiency.

“I genuinely believe that when there is tremendous demand for certain things, it will also lead to supply coming through,” he said.

As those constraints ease, AI infrastructure should become more affordable for enterprises, supporting wider adoption across the economy.