Malaysia’s data-centre electricity consumption is projected to surge from 10,544 gigawatt-hours in 2026 to 73,274 gigawatt-hours by 2035, according to a July 16 report, as AI demand outpaces clean energy deployment. The gap is being filled by coal, which remains the fastest available firm power source in a region that holds a majority share of global reserves.
The buildout is financed in part by Western hyperscalers — Google, Microsoft, and Amazon have committed a combined US$10.4 billion in Malaysia — whose climate pledges are colliding with the energy reality on the ground.
The cloud services that power Western AI applications are increasingly running on coal. In Malaysia, where Google, Microsoft, and Amazon have together pledged US$10.4 billion, the electricity that keeps server farms running is coming from a grid still dominated by fossil fuels — and the fastest way to add more capacity is to burn more coal. The contradiction is not hidden. It is embedded in the infrastructure itself.
For Adit Rahim, a 49-year-old communications executive living in Cyberjaya, the data centre boom is not an abstraction. He says the strain on local utilities will be very noticeable. His concern is the human edge of a much larger tension: the speed at which AI is expanding is forcing governments and developers to choose between reliability and emissions, and reliability is winning.
The grid was never built for this
“Malaysia’s grid was not designed for the reliability and continuous baseload power that hyperscale AI campuses need,” said Nik Nur Zafirah, senior manager at PwC Malaysia. “What that really means is continuous baseload power.” The country’s data-centre pipeline in Johor alone could reach 2,055 megawatts by the end of 2026, with another 3,500 megawatts beyond that, according to property consultancy JLL. Each new megawatt of server capacity demands a megawatt of firm electricity, and the queue is growing faster than the supply of clean alternatives.
Tenaga Nasional, the national utility, has committed RM43 billion between 2025 and 2027 to strengthen the grid. But grid upgrades take years, and the projects already approved need power now. The result is a widening gap between what the system can deliver and what hyperscalers are asking for. Alvin Tan, a power and renewables analyst at Wood Mackenzie, noted that the main constraint is shifting from generation capacity to where power can actually be delivered, making grid readiness the limiting factor.
The tension between AI expansion and clean energy timelines is easier to see than to describe.
| Country | Current rule | New rule | Effective date |
|---|---|---|---|
| Malaysia | No mandatory pre-approval capacity assessment for data centres | Data Centre Task Force must check power and water capacity before approvals | July 2026 |
| Singapore | Data centres capped at ~12% of national grid load | Second application round (Dec 2025) requires at least 50% green power | December 2025 |
| Malaysia | Grid investment not specifically tied to data centre demand | TNB’s RM43 billion grid upgrade plan (2025-2027) targets data centre growth | 2025–2027 |
Malaysia’s clearest response is procedural rather than punitive. The Data Centre Task Force, announced by Deputy Investment, Trade and Industry Minister Sim Tze Tzin in Parliament on July 16, now checks power and water capacity before approvals. The government says excess capacity must exist before projects proceed. Singapore is using a tighter model, capping data centres at around 12% of grid load and requiring green-power participation in its December 2025 application round. Both approaches still manage demand more than they transform the supply mix quickly enough.
The emissions embedded in a ChatGPT query processed in Johor do not appear on any Western balance sheet, but they accumulate in the atmosphere just the same. Whether coal becomes the default power source for Asia’s AI buildout is not a settled policy choice, but analysts and industry reports point to it as the path of least resistance.
The speed premium that coal buys
The financial driver is not just cheap electricity. It is the speed at which capacity can be delivered. Hyperscalers and developers pay a premium for immediate firm power, and that shifts bargaining power toward utilities and grid owners who can offer it first. In a region where coal plants can be built or restarted faster than large-scale solar and wind farms can be commissioned and connected, the incentive structure points toward coal.
Asia controls nearly three-fifths of the world’s coal reserves, by some estimates. That endowment makes coal a low-cost, geopolitically insulated option compared with imported oil and gas. For a data-centre operator facing a deadline, the choice is not between coal and renewables in the abstract. It is between a coal plant that can deliver power in months and a solar farm that will take years to clear permitting and grid connection.
Watch for Malaysia’s 2025-2027 grid-upgrade execution and new data-centre approvals under the Data Centre Task Force. If they accelerate, it means regulators believe supply can catch up with AI demand. If they stall, more projects will shift toward whatever firm power is available first. Singapore’s next low-carbon data-centre allocation round, expected in 2026, will signal whether the region is still rationing clean capacity or beginning to expand it. The next twelve months will show whether the pattern breaks — or confirms itself for another cycle.
Beyond the headline
The money trail
The rapid buildout is driven by the financial imperative to secure capacity immediately, not just by low electricity prices. Developers and hyperscalers are paying for speed, and that concentrates power among utilities and landowners who can deliver firm electricity first. The long-run climate cost is higher, but the short-run calculus rewards whoever can connect the fastest.
The reach
Western firms that outsource AI workloads to Malaysia inherit the emissions embedded in those digital supply chains. Climate pledges made at headquarters become harder to defend when the computing is powered by coal abroad. The reputational risk is not theoretical — it is already priced into the next sustainability report.
What isn’t being said
The dominant framing treats this as an Asian energy shortage, but the missing layer is the influence of global tech buyers. If those buyers insisted on firm low-carbon power as a condition of siting, the development pattern would change faster than if governments alone are asked to catch up. The demand signal from the world’s largest cloud customers could reshape the energy mix more quickly than any regulation.
The choices that will define the next buildout
With Malaysia’s data-centre pipeline accelerating and Singapore’s next green-power allocation round approaching, four groups face decisions that will shape the carbon footprint of Asia’s AI infrastructure.
- Western Tech Executive with APAC Data Center Operations
You need to assess the energy mix of your current and planned data centre infrastructure in Malaysia and other APAC countries. Review the latest sustainability disclosures from your cloud providers and ask whether the power purchase agreements behind your Johor capacity are backed by firm renewables or by coal-heavy grid supply. The reputational and regulatory risk is material, and the next reporting cycle will make it visible.
- ESG Investor with Exposure to Asian Utilities or Tech Infrastructure
Re-evaluate your portfolio’s exposure to companies involved in Asia’s data centre buildout. Scrutinise the energy sourcing and transition plans of utilities like Tenaga Nasional and developers active in Johor. The gap between stated climate targets and the coal-dependent reality is a stranded-asset risk that will widen if clean capacity does not catch up.
- Malaysian Government Official for Energy or Investment
The Data Centre Task Force is a start, but it manages demand rather than accelerating supply. Consider policy adjustments that tie approval timelines to firm renewable capacity milestones, and prioritise grid interconnection for solar and wind projects in data centre corridors. The RM43 billion grid upgrade is necessary but not sufficient without faster low-carbon generation.
- Global Supply Chain Manager for Semiconductor or IT Hardware
Investigate the energy consumption and sourcing practices of your downstream partners and cloud service providers in Asia. The embedded carbon in your IT supply chain is increasingly tied to the coal that powers data centres in Malaysia. Request energy attribution data from your hyperscale vendors and factor it into your Scope 3 accounting before the next audit cycle.
Explainer
- Hyperscale data centre
- A massive, highly efficient computing facility designed to scale rapidly to meet the demands of large cloud and AI workloads. Hyperscale operators like Google, Microsoft, and Amazon run tens of thousands of servers in a single campus, requiring continuous, high-reliability power. The term reflects not just size but the ability to add capacity in large, modular increments.
- Baseload power
- The minimum level of electricity demand that must be met continuously, 24 hours a day. For data centres, baseload power must be firm and uninterrupted, which is why intermittent renewables like solar and wind cannot yet fully replace coal or gas without large-scale storage. The reliability requirement is what makes coal an attractive short-term option for AI infrastructure.
- Data Centre Task Force
- A Malaysian government body established to assess power and water supply capacity before approving new data-centre projects. Announced in Parliament on July 16, 2026, it reflects a policy shift from open-ended development to conditional approval. The task force’s effectiveness will depend on whether it can accelerate clean energy supply rather than simply slow down demand.
- National Energy Transition Roadmap
- Malaysia’s long-term plan to reshape its energy mix, targeting 70% renewable generation capacity by 2050, up from about 39 GW in 2023 to 97 GW. The roadmap sets ambitious goals but faces a timing mismatch: data-centre demand is surging now, while the renewable buildout is scheduled over decades. That gap is what coal is filling in the interim.
- Firm power
- Electricity that is available on demand, regardless of weather or time of day, and can be dispatched instantly to meet load. Coal, natural gas, nuclear, and hydro with storage all provide firm power. In the context of AI data centres, firm power is the non-negotiable requirement that forces developers toward fossil fuels when clean firm options are not yet built.
RELATED STORIES
