Malaysia's Data Centre Boom

Analysts and industry bodies rank Malaysia among the fastest growing data centre markets in Southeast Asia. According to the Ministry of Investment, Trade and Industry (MITI), the country approved roughly RM144.4 billion (approximately US$31 to 32 billion) across 143 data centre projects between 2021 and June 2025. These figures are approved rather than built or spent, and different sources count facilities differently: one tally cites around 34 operational data centres in early 2026, with dozens more under construction and one broader count of about 123 facilities across all stages.

The timing has three drivers. First, cloud demand from hyperscalers matured. Second, the generative AI wave sharply raised demand for high density compute, which needs large, power dense sites. Third, and most specific to Malaysia, Singapore constrained its own data centre growth from around 2019, redirecting nearby demand across the Johor Strait.

The result is a large, front loaded pipeline. Data centre electricity draw reached about 603 MW as of June 2025, roughly 47% of a declared maximum demand of about 1,276 MW, per grid operator and Energy Commission figures. Proponents (MITI, MIDA, operators) frame this as strategic digital infrastructure and foreign investment. Skeptics (civil society groups, some academics and journalists) argue the sector consumes disproportionate power and water for relatively few permanent jobs. This guide sets out both positions using the same underlying numbers, and flags where the evidence is contested.

Four factors explain the concentration of investment.

Singapore spillover. Singapore imposed an effective moratorium on new data centres from around 2019 amid land, power and carbon constraints. It lifted the pause selectively in 2022 with a pilot allocating only a small tranche of new capacity (approximately 80 MW initially), keeping Singapore supply tight and pushing growth toward Johor and the Klang Valley.

Cheaper power. Before the 2025 tariff changes, Malaysian industrial tariffs (around US$0.10/kWh) ran at roughly one third of Singapore's (around US$0.27/kWh). This gap was a central part of the pitch, and part of what the July 2025 tariff restructuring narrowed.

Land and connectivity. Malaysia offers abundant, cheaper industrial land, subsea cable landings, and low latency proximity to Singapore's connectivity ecosystem.

The JS-SEZ. The Johor-Singapore Special Economic Zone streamlines cross border trade and investment and aims to integrate power and connectivity, with the digital economy among its flagship sectors. The two governments signed a memorandum of understanding in January 2024 and a full agreement in January 2025.

Taken together, these advantages let operators pair Singapore capital and demand with Johor land and power. Analysts note the advantages are real yet time-limited: the power cost gap has narrowed, water rules have tightened, and grid capacity is finite, so the durability of Malaysia's edge is now part of the debate.

The investor base spans hyperscalers, global and regional operators, Chinese platforms, and Malaysian government linked companies. The figures below mix multi year cloud spend with physical data centre capital expenditure and are not directly comparable. 'Not disclosed' means the amount is not separately or publicly itemised.

In September 2024 Blackstone led a deal valuing AirTrunk at roughly US$16 billion, a signal of global capital appetite for Asia Pacific, including Johor, capacity.

Capacity is concentrated in the south, with a secondary cluster in the Klang Valley and emerging northern sites. Johor is the clear epicentre and is expected to hold around 60% of national capacity by 2030 (approximate). The state's live supply grew very rapidly off a small base between 2019 and 2024.

The southern concentration is deliberate: it maximises proximity to Singapore and the JS-SEZ. It also concentrates resource and grid exposure. A few very large loads sitting on specific state grid segments raise local reliability and water supply questions that are less acute when capacity is spread across a country. That concentration is why Johor, rather than Malaysia as a whole, sits at the centre of the electricity and water debates that follow.

Johor illustrates the boom's promise and its pressure points in one place. Southern Johor became the natural overflow when operators sought nearby land with cheaper power and low latency to Singapore, and AI demand later amplified the trend. Estimates vary, but Johor alone is often cited with a committed and planned pipeline in the range of roughly 5 GW or more, while operational capacity is a much smaller fraction (very roughly 1 to 2 GW as of 2025 to 2026).

The main clusters sit around Kulai and Sedenak, including Sedenak Tech Park and YTL's Green Data Center Park, plus Nusajaya and Iskandar Puteri. Individual early phase campuses are large: Princeton Digital Group and AirTrunk are each cited at around 150 MW early phase, and Yondr at around 200 to 300 MW. The YTL and Nvidia campus is planned at up to around 500 MW in total, though its first Nvidia powered facility, completed in October 2025, is about 20 MW.

The strain shows up in water. In November 2025, Johor authorities asked investors to postpone water cooled expansions for about 18 months (to roughly mid 2027) amid drought, and halted new approvals for the largest, highest water use facilities (often described as Tier 1 and Tier 2). Reporting notes acute supply and demand gaps in parts of the state, which also supplies raw water to Singapore under a 1962 agreement. Johor therefore functions as the live test of whether Malaysia can host the sector while protecting domestic power and water supply.

The single most useful snapshot is approved capacity against actual usage. In early 2026, government figures (via PETRA) showed actual consumption running at roughly half of what has been approved, for both power and water.

The government reads this table as headroom: usage is about half of approvals, so the system is not yet strained. Critics read the same table forward: the applied for pipeline (reportedly above 11,000 MW) is many times current usage, so the concern is about projected future load rather than present load. Both readings rest on the same figures and differ on how to weight present headroom against projected demand. Note that the two approved versus actual pairs come from different sources: the June 2025 draw of about 603 MW against a declared maximum demand of about 1,276 MW is a grid operator and Energy Commission figure, while the early 2026 use of about 1,102 MW against about 2,050 MW approved maximum demand is from PETRA. They use different denominators (declared versus approved maximum demand), so they are point in time snapshots that are not directly comparable rather than a single continuous trend.

Electricity pricing is the most emotionally charged issue, because households worry the boom will raise their bills.

Effective 1 July 2025, Malaysia implemented a restructured tariff under the regulatory period commonly called RP4 (covering roughly 2025 to 2027). It revised the base tariff and introduced more granular components such as time of use and demand based charges, intended to make heavy users pay according to their consumption patterns. The average base tariff rose about 14% (from roughly 39.95 to 45.62 sen/kWh, effective 1 July 2025 to end 2027), a general change that applies to all Peninsular Malaysia users rather than a data centre specific surcharge. Separately, analysts estimate data centre operators could see power costs rise by around 10 to 14% under the restructured framework.

The two framings:

• Government and TNB: tariffs are being designed so large users, including data centres, bear their fair share of grid and infrastructure costs rather than shifting them to households. This is presented as cost reflective pricing.
• Critics (for example commentator Murray Hunter and some think tanks): the concern is whether ordinary domestic consumers might indirectly bear grid expansion and subsidy costs over time, and whether cheap power incentives used to attract data centres are sustainable. Some operators, separately, view the increase as eroding Malaysia's cost advantage over Singapore.

The net effect on ordinary consumers remains debated. The tariff restructuring narrows the power cost gap with Singapore that helped attract operators in the first place, which is why pricing sits at the centre of both the civic anxiety and the competitiveness question.

Rapid load growth raises the question of whether generation and transmission can keep pace. Peninsular Malaysia's grid has a peak demand of roughly 20 GW, so multi gigawatt data centre commitments over the coming decade are a material share.

TNB and the Energy Commission have introduced fast track connection mechanisms (see the Green Lane Pathway) and are expanding transmission and generation. Officials express confidence the grid can absorb planned load if it is managed. The concerns raised by MIDA, Bernama and think tanks such as Fulcrum (ISEAS) are threefold:

• Generation adequacy and gas dependence. Malaysia's grid remains heavily gas and coal fired. Large new baseload from data centres can lock in fossil generation and raise system emissions, even as operators buy green attributes on paper. Analysts question whether clean energy build out can keep pace.
• Feedstock and upgrades. New supply leans on natural gas, raising questions about gas feedstock availability and the cost of grid upgrades.
• Load concentration. A few very large loads in specific states, notably Johor, concentrate risk on local grid segments, so the risk is driven by where the load sits rather than by national totals.

A recurring caution from analysts is that connection commitments can outpace actual generation build out. In other words, signing up load is faster than building the plants and lines to serve it, which is why grid planning has become a national issue rather than a utility one.

Data centres that use evaporative or water based cooling can consume from hundreds of thousands to several million litres per day at large scale. A common rule of thumb is around 4 million litres per day to cool a roughly 100 MW facility, and press accounts describe large AI campuses using water on the order of Olympic pools per day.

This matters most in Johor, a water stressed state that also supplies raw water to Singapore under a 1962 agreement. In November 2025, Johor asked investors to postpone water cooled expansions for about 18 months and halted new approvals for the largest, highest water use facilities.

The policy direction is to shift industrial cooling off potable water:

• Draft National Water Reclamation Policy (rolling out in 2026, per Deputy PM and PETRA Minister Fadillah Yusof): route reclaimed and recycled wastewater to industrial users including data centres, reserving treated potable water for households. Output target: from about 48.5 MLD today to about 118 MLD by 2030.
• Johor operators are already shifting toward reclaimed water under tighter state rules and higher costs.

The scale gap is the risk. Malaysia's reclaimed output (about 48.5 MLD now) sits far below Singapore's roughly 760 MLD, while three states are projected to need around 445 MLD. Until reclamation scales, evaporative cooling competes with household and agricultural demand, and switching to less water intensive cooling tends to raise energy use instead.

The tension is between the boom and Malaysia's decarbonisation goals, set out in the National Energy Transition Roadmap (NETR, launched August 2023).

Operators can procure renewable power through two main instruments:

• CRESS (Corporate Renewable Energy Supply Scheme, launched September 2024): lets large consumers buy green power directly from developers via the grid, paying a system access or wheeling charge.
• GET (Green Electricity Tariff): a subscription to renewable energy.

Solar, both utility scale and rooftop, is the main domestic renewable source. Operators also advertise low PUE (power usage effectiveness) and renewable procurement.

The contested point is additionality. Green procurement (CRESS, GET and solar power purchase agreements) can decouple operators from grid emissions on paper, but the underlying grid mix remains gas and coal heavy, which limits how green the sector currently is. Analysts such as ISEAS and TransitionZero question whether clean energy build out can keep pace with demand.

On the framework side, PETRA and MIDA are developing sustainability guidelines and a net zero oriented energy and water supply framework for the sector, including efficiency expectations (PUE and WUE) and e waste concerns. Guidelines on Sustainable Data Centres emerged in 2024, tying incentives to efficiency and sustainability criteria and signalling a shift from attracting volume toward managing intensity. Whether these measures offset projected load in time is the crux of the sustainability debate, and both proponents and critics agree the outcome depends on renewable and reclaimed water build out speed.

The employment case is genuinely two sided, and both sides are supported by the same facts.

The upside. Data centres attract large foreign direct investment, build strategic digital infrastructure, and generate significant construction and engineering work during the build phase. They support indirect roles in power, cooling, security and maintenance, plus higher skilled operational and cloud jobs. The Asia-Pacific Data Centre Association (APDCA) estimates, in a scenario projection, that the Malaysian AI and data centre industry could create on the order of about 30,900 jobs annually by around 2030, up from about 4,400 in 2024, alongside an estimated total economic output of around US$34 billion.

The caveat. Data centres are capital intensive and highly automated, so direct permanent operational headcount per facility is modest. A frequently cited example: the 25 projects holding Malaysia Digital (MD) status with DESAC incentives are projected to create about 1,429 jobs, a figure critics note is small relative to the capital deployed. Much job creation sits in construction and indirect or induced activity rather than long term, high density employment, so job per megawatt and job per ringgit returns are debated.

The balanced read is that data centres deliver capital, infrastructure and a burst of construction employment, alongside far fewer permanent jobs than comparable manufacturing investment. Whether that trade is worthwhile depends on how much weight is placed on digital infrastructure and FDI versus direct long term employment, which is a value judgment rather than a factual dispute.

Governance has tightened as the boom has scaled. The government's stance has moved from attracting volume toward matching each project to available resources.

A national task force (involving PETRA and state governments) is reviewing data centre projects to ensure each has matching energy and water availability. The principle is that projects are approved only where sufficient power and (reclaimed) water capacity exist, with the framework cascaded to the state level.

The main instruments in play:

• Green Lane Pathway (TNB, introduced around 2023): fast tracks grid connection for qualifying data centres, cutting energization from roughly 36 to 48 months to as little as about 12 months, plus a one stop centre for investors.
• CRESS and GET: renewable procurement routes.
• Draft National Water Reclamation Policy: water reuse, reserving potable water for households.
• RP4 tariff: cost reflective price signals, effective July 2025.
• Sustainable data centre guidelines: efficiency, emissions and e waste expectations.

In November 2025 the government also announced a proposed single licensing authority to streamline and better regulate the sector, consolidating what has been a fragmented approval landscape. Taken together, the policy trend is toward conditional approval: welcome investment, but only where power and water can be shown to exist, and price the externalities through tariffs and water rules rather than leaving them uncosted.

Beyond power, water and tariffs, three structural risks recur in analyst and civil society commentary.

Obsolescence and stranded assets. AI hardware cycles are short. If AI demand shifts or chips are superseded, some capacity, especially AI specific hardware, could become obsolete, leaving energy and water commitments without matching economic return.

Concentration. Heavy geographic concentration in Johor amplifies resource and grid exposure, so a local water or grid constraint has outsized national effect.

Bubble risk. Some analysts caution that a portion of the pipeline is speculative, announced or land banked rather than built, and that slowing AI demand, hardware obsolescence, or rising energy and financing costs could leave excess capacity. Others argue underlying cloud and AI demand is structural and Malaysia's cost and location advantages are durable. The outcome is genuinely uncertain.

The balanced bottom line. Proponents (MITI, MIDA, operators) emphasise investment, digital infrastructure sovereignty, construction and skilled operations jobs, and positioning Malaysia as an AI backbone for Southeast Asia. Skeptics (civil society groups such as Aliran, and commentators including Murray Hunter, plus think tanks like ISEAS and TransitionZero) counter that per ringgit job creation is low and that power and water externalities are under priced. The government position is that current usage is about half of approved capacity, so there is headroom and no present crisis. Both positions rest on the same figures. They differ on how to weight present headroom against projected load, so the disagreement is about how to weight the evidence rather than about the evidence itself.