AI boom casts doubt on ASEAN cross-border power trading

Key points:

• Rapidly rising energy demand from data centres has made the Thailand-Malaysia power-sharing expansion unfeasible, casting doubt on wider plans across the region.

• 24/7 AI and data centre loads have eliminated the offset peak hours that previously enabled cross-border energy trading.

• Regional interconnections will shift from routine energy trading to emergency tools for grid stabilisation.

---

A planned power-sharing project with Malaysia is “no longer feasible” since data centres have driven up domestic energy demand, according to a senior figure at Thailand’s state utility.

Current demand means the expansion of the Thailand-Malaysia grid interconnection may not go ahead, warned Warit Rattanachuen, deputy governor for strategy at the Electricity Generating Authority of Thailand (EGAT).

Rattanachuen said preliminary simulations showed that both countries increasingly need the electricity they each generate for their own needs, limiting the scope for cross-border power trading. He was speaking at Enlit Asia in Jakarta last week.

The warning comes as Asean governments push ahead with plans to connect their electricity systems more closely, allowing countries to share power and renewable energy across borders. The Thailand-Malaysia link is a longstanding part of that effort: the existing 300MW connection has operated since 2002, and is expected to end in 2027.

But the growth of data centres is changing the way electricity demand is distributed across the two countries.

Data centres consume large amounts of power around the clock, and with Malaysia and southern Thailand now having similar energy usage, power sharing has become less viable than before, Rattanachuen said.

“In the past, Malaysia might have peak demand in the afternoon and southern Thailand may have peak demand in the evening, so we can share resources. But now, when we run our simulations, we don’t have enough energy for the new [digital] infrastructure,” he said. “Every country in Asean needs the power supply and renewable energy, especially solar.”

The shift comes as Southeast Asia’s data centre capacity is expected to triple from 2025 levels by 2030, with Malaysia and Thailand among the fastest growing markets.

But mounting concerns over resource consumption and local community impacts have led Thailand to recently suspend 49 new data centres to give officials time to develop new industry regulations, while Malaysia saw its first data centre protest in the southern state of Johor earlier this year.

Accurately forecasting electricity demand is now a challenge for utilities, as data centre capacity increases rapidly year-on-year, added Rattanachuen.

Three years ago, data centres were estimated to require around 30-50MW of power, but facilities now regularly exceed 1GW in capacity, especially for those serving AI computing needs, he said.

As a result, many developers now “just want firm energy and do not care whether or not it is green,” Rattanachuen said.

The surge in energy demand from data centres and electric vehicles is also creating new challenges on the supply side, he said, making grid flexibility and resilience increasingly important.

In order for new cross-border connections to remain viable, policies will need to ensure they provide regional energy security value, Rattanachuen told Green Central Banking. This could include reducing the risk of power outages and creating opportunities for future Asean collaboration, he said.

But the changing role of cross-border links may not be limited to power trading.

Phuwanart Choonhapran, assistant secretary general at Thailand’s Energy Regulatory Commission, said that with AI, the value of regional interconnections, such as the links Thailand shares with Laos and Malaysia, will come from being able to stabilise grids when there are load fluctuations.

“These two main interconnections serve both normal operations and emergency operations,” said Choonhapran. “We will see more emergency operations in the near future because we’re going to have large loads from the AI data centres – and when they trip on the grid, they lead to frequency swings.”

 

Written by