KUALA LUMPUR: Prime Minister Anwar Ibrahim confirmed in parliament on 24 February 2026 that applications for data centres unrelated to high technology and AI had been stopped, formalising a policy that had in practice been running since the middle of 2024. The informal moratorium was never publicly announced when it began, according to a report by Tech Wire Asia. The confirmation makes explicit what developers had been working around for more than a year.
The numbers behind the decision are large. Malaysia's data centre capacity grew from about 10MW in 2021 to about 1.3GW in 2024, and Johor is expected to account for 60 per cent of national capacity by 2030. According to DC Byte, more than two-thirds of the capacity under construction across Southeast Asia's five largest economies sits in Malaysia. Major projects include the USD4.3 billion AI data centre run by Nvidia and YTL, a second land purchase by Microsoft and a USD2.1 billion expansion by ByteDance.
Utilities, rather than land or capital, set the pace. Johor rejected about 30 per cent of new data centre applications in 2024, citing utility timelines that did not match project schedules and planning that arrived too late in the process. Some power and water approvals take up to 18 months, which means a developer can hold land and capital while waiting for a connection.
A moratorium aimed at quality, not volume
The distinction the government draws is between data centres tied to high technology and AI and those that are not. A facility built for general-purpose colocation places a similar load on the grid and the water network as an AI campus, but supports a different kind of economic activity. By stopping approvals in the second category, the policy steers scarce power and water towards the projects the government considers strategic.
- Applications for non-AI data centres stopped
- Informal moratorium running since mid-2024
- Johor rejected about 30 per cent of applications in 2024
- Some utility approvals take up to 18 months
Parliamentary data from November 2025 showed a separate problem on the demand side. Malaysia's data centres were consuming only 603MW against a declared maximum demand of 1,276MW, a utilisation rate of roughly 40 per cent, while the government's target is 85 per cent. Declared maximum demand is the capacity an operator reserves, so the gap between what is reserved and what is drawn has direct consequences for grid planning.
Who pays for the grid
Then deputy energy minister Akmal Nasrullah said in December 2025 that the cost of grid upgrades tied to data centres must be borne in full by developers and must not be passed on to electricity users as a whole. The principle matters because data centre load is concentrated in a few corridors, while the transmission work needed to serve it is not. Passing those costs to all users would spread the benefit of a private project across the public.
- 603MW consumed against 1,276MW declared maximum demand
- Roughly 40 per cent utilisation
- Government target of 85 per cent use
- Grid upgrade costs to fall on developers
Prime Minister Anwar Ibrahim confirmed as much in parliament on February 24, telling lawmakers that applications for data centres unrelated to high-technology and AI had already been stopped.
EY Asia-Pacific energy and resources leader Mark Bennett estimates that data centres alone will need 5-6GW of power by 2035, roughly a fifth of Peninsular Malaysia's current total generation capacity. To meet that demand and decarbonise at the same time, Malaysia plans to add 6-8GW of gas-fired generation by 2030 and up to 10GW of renewable energy. Those additions have to be built, connected and financed within the same window.
Sovereign cloud and the money behind it
The government has committed RM2 billion to build a sovereign AI cloud, part of a broader RM5.9 billion commitment to AI research, development and commercialisation. Sovereign cloud in this context means infrastructure the state controls for public sector workloads rather than capacity rented from commercial providers. The commitment ties the moratorium to a build-out, since restricting approvals for non-AI projects only makes sense if the state is also creating the capacity it wants used.
Parliamentary data from November 2025 showed that Malaysia's data centres were consuming only 603 megawatts against a declared maximum demand of 1,276 megawatts – roughly 40%. The government's target is 85% use.
What Ireland shows about the end of the road
Ireland offers a cautionary comparison. In 2024 data centres accounted for 22 per cent of the country's metered electricity, more than urban households, and the grid operator EirGrid has restricted new connections in Dublin. Malaysia is not close to that share today, but the concentration of load in Johor and the long lead times on grid work mean the constraint can appear faster than national averages suggest.
For operators, the practical consequence is that approvals now depend on how a project is classified as much as on its credentials as a customer. Projects that cannot be described as AI or high technology face a much harder route, while those that can still have to answer questions about power, water and the timing of their connections.
- Classify the project correctly before applying
- Expect utility timelines to set the schedule
- Budget for grid upgrade costs
- Plan for water as well as power
For Malaysia's cloud and data centre market, the moratorium reframes growth around quality rather than headline capacity. Capacity already committed will still be built, but the pipeline for general-purpose colocation will be thinner than it would have been. The utilisation gap, the grid cost principle and the Irish precedent all point the same way: the next phase of Malaysian data centre growth will be decided by power, water and policy rather than by land availability alone.
Source: Tech Wire Asia
