Malaysia's Data Centre Surge Is Swallowing a Third of the Grid, and Ratepayers Aren't Being Told Who Pays

A report from Eco-Business last week documented Malaysia's power grid straining under simultaneous data-centre expansion and renewable integration.^[1] The headline facts are stark: 36 operating data centres drawing 4.5 GW of capacity through first quarter 2026, another 23 projects under construction carrying 3.8 GW, and 79 more in planning stages with roughly 25.6 GW.^[7] Data centres now account for 9.28% of Malaysia's total electricity consumption, a share expected to climb to 31% by 2035.^[6]

But the urgent question buried beneath the growth numbers is this: who absorbs the cost if these announced pipelines underdeliver? Tenaga Nasional Berhad (TNB), Malaysia's national utility, has committed 43 billion ringgit (about $10.85 billion USD) in grid modernization through 2027.^[8] That capital is being raised through rate base recovery, which means residential and small-commercial customers are funding transmission and distribution upgrades justified by data-centre forecasts that may never materialize at the scale promised. The mechanism is familiar from North American practice: utilities compress connection timelines (TNB's Green Lane Pathway compresses the usual 36, 48 months to 12 months^[9]), front-load capex against signed contracts, then lock the cost into rates regardless of whether the load shows up. The first sign of a lid on the boom, Malaysia's government pause on non-AI data centres citing water and power constraints^[3], should trigger immediate disclosure of (1) which of TNB's 3.8 GW under-construction projects carry binding minimum-demand ratchets (what share of contracted capacity must the customer pay for, whether or not it draws power?); (2) contract term lengths against the 40+ year life of the transmission assets being built; and (3) whether costs are socialized across all customer classes or isolated to a separate large-load tariff. TNB's most recent parliamentary filing states peak demand is forecast to rise from 21.3 GW in 2026 to 33.5 GW by 2035,^[7] with data centres driving nearly all of that growth. No disclosure appears to exist, in any docket or public filing, of the collateral, exit fees, or cost-responsibility language in the individual customer contracts.

This matters because the contract structure determines whose balance sheet absorbs a shortfall. In jurisdictions with strong large-load tariffs, Virginia's GS-5, Ohio's AEP agreement, Oregon's Schedule 96, the customer class itself carries the cost of reserved capacity and unamortized plant if demand falls short. Minimum-demand ratchets (typically 60, 85% of contracted generation and transmission over 10, 14+ year terms) ensure that unused infrastructure isn't dumped on residential rates. Malaysia has published no equivalent tariff framework. TNB operates under cost-of-service regulation, meaning the utility recovers approved capex plus a return regardless of utilization. For a utility, that is the ideal structure: forecast high, build fast, lock the cost in. For ratepayers, it is a bet they cannot audit.

The government pause on non-AI data-centre builds, framed as temporary, driven by water and power resource limits, is a pressure release valve, not a reset. It does not address the underlying problem: TNB's capex authorization and the contracts already signed operate in isolation from public tariff disclosure. The solution is straightforward and exists in prototype form: (1) require all new data-centre service agreements to specify minimum-demand ratchets of at least 80% over their contract term, collateral of at least $1, 1.5 million per MW to cover stranded-cost risk, and exit fees covering unamortized investment; (2) isolate the cost of dedicated network upgrades to a separate large-load customer class, as Virginia, Ohio, and Oregon have done; (3) mandate that any data-centre contract term be matched to the asset life of the generation and transmission it requires, so 15-year demand commitments do not underpin 40-year plants; and (4) publish all non-confidential terms, demand charge, ratchet percentage, collateral, term, cost-allocation mechanism, so the public can assess stranded-cost exposure. The window to impose these rules before more capacity is signed is closing fast. Once contracts are inked and capex is approved, the cost is locked and ratepayers have no leverage.