Malaysia’s rapid build-out of artificial intelligence and data centre infrastructure is set to place increasing pressure on electricity supply, potentially extending the role of gas and other dispatchable generation as the country balances investment growth with its energy-transition ambitions, according to BMI.
In its latest Asia power-sector assessment, BMI forecasts Malaysia’s electricity consumption to grow by an average 3.3% annually over the next decade, alongside 2.9% growth in Thailand and 5.6% in Vietnam.
The research house said data centre investment and expanding AI workloads were behind recent upward revisions to its power-demand forecasts. It estimated that data centre capacity in Malaysia will increase by around 50% year-on-year to 1.6GW in 2026.
Malaysia’s data centre expansion is already becoming a significant component of national electricity demand. The Energy Transition and Water Transformation Ministry said in July that TNB was supplying 36 operating data centres with planned capacity of about 4.5GW, while another 23 projects under construction had maximum demand of 3.8GW. The ministry projects data centres could account for as much as 31% of electricity consumption by 2035, compared with around 7% in 2026.
BMI said the pressure could be greater than conventional capacity forecasts suggest because hyperscalers are increasingly building their own facilities, which are not fully captured in traditional colocation data centre projections.
The research house expects Malaysia’s electricity supply constraints to peak in 2026 before additional generating capacity provides some relief in 2027. However, continued hyperscaler expansion could keep the market tight for longer.
BMI said data centres accounted for a record 9.3% of Malaysia’s electricity consumption during the second week of August, compared with an average of about 7% during the year, with hotter weather further increasing cooling demand.
Tighter Scrutiny On Data Centre Expansion
Regulatory scrutiny is also increasing as governments seek to prevent large new loads from compromising grid reliability and water availability.
BMI said Malaysia is tightening conditions around data centre connections, with operators increasingly expected to demonstrate adequate power and water arrangements.
The government has separately said new data centre projects will only be approved after authorities establish that electricity and water capacity is sufficient, with domestic users and existing industries remaining a priority.
Malaysia has also sought to prevent infrastructure costs associated with the sector from being passed on to ordinary consumers. The government previously said grid and water infrastructure upgrades required specifically for data centres would be borne by developers.
The current Regulatory Period 4 framework, running from July 2025 through December 2027, also provides for significant electricity-network investment as the system prepares for higher demand.
Gas And Coal Could Gain From AI Power Demand
BMI expects the increase in electricity consumption to strengthen the case for dispatchable thermal generation, particularly in emerging Asian markets where grids need generation that can respond quickly and provide continuous supply.
Coal and gas remain potential short-term solutions because renewable capacity alone may not provide the firm power required by rapidly expanding data centre loads without sufficient storage, transmission and other grid infrastructure, BMI said.
The research house expects this tension between AI-led economic investment and decarbonisation objectives to become increasingly important over the next decade.
BMI’s view is that emerging economies could prioritise ensuring adequate electricity supply for investment and economic growth even where that results in a longer role for thermal generation.
Malaysia, however, is also preparing substantial grid investment for the increase in power demand. TNB’s RP4 framework includes RM16.27 billion of contingent capital expenditure, which may be triggered by additional demand including data centres and energy-transition projects.
BMI said the fundamental issue is the difference in development timelines: large AI and data centre projects can add electricity demand much faster than new transmission infrastructure can be planned and completed.
As hyperscaler investment continues, that gap is expected to make grid capacity, access to firm electricity and power costs increasingly important factors determining where new Asian data centres are built, rather than land availability and connectivity alone.
