The Grid as a Strategic Asset: Unlocking Transmission Flexibility in Southeast Asia

The first piece in this series argued that data centers are not merely passive consumers of electricity but potential anchors of the low-carbon energy infrastructure of the future. That argument rested on the observation that the energy demand of hyperscale facilities is large, steady, creditworthy, and increasingly shaped by zero-carbon sourcing requirements, making it structurally unlike any other load on a modern power system. The question that follows is whether the planning, regulatory and institutional architecture of the energy systems receiving that demand is actually designed to capture its value. In most of the world, and particularly across Southeast Asia, the answer is not yet.

The region is experiencing a data center investment surge of heightened scale and speed. The six major ASEAN economies have collectively emerged as the next global data center hotspot, with 2.9 GW of total capacity currently in the pipeline. Malaysia has attracted substantial data center and cloud investment since 2021, with MIDA reporting RM184.7 billion directed to data center and cloud projects as part of RM278 billion in total approved digital investment through 2024, across 143 approved data center projects as of mid-2025. Data center power demand in Malaysia is projected to reach 68 TWh by 2030, accounting for roughly 30% of the country’s total national electricity consumption at that point. These represent a structural transformation of electricity demand in a region where grids remain heavily dependent on fossil fuels and where the regulatory frameworks governing how large loads participate in system planning were not designed with this kind of demand in mind. The choices made now about whether this demand is treated as a passive load or an active system resource will shape the region’s energy profile for decades.

Here, we focus on what those choices look like in practice, with Malaysia as the central case study. We argue that the dominant policy conversation in the region remains oriented toward supply expansion and cost allocation, and that a more consequential and underexplored opportunity lies in reframing large data center loads as active participants in grid planning and operations. The institutional, planning and regulatory shifts required to realize that reframing are tractable, and Malaysia’s existing policy momentum places it in a position to lead on them.

What Flexibility Actually Means

Data centers possess meaningful and largely untapped capacity to modulate their electricity consumption in ways that reduce system costs, facilitate renewable integration, and substitute for some portion of firm thermal generation capacity. This constitutes system flexibility: as controllable loads, they can shift or curtail demand to support grid stability and reduce the need for additional generation and network capacity. The IEA has estimated that if data centers offer as little as 30 hours of flexibility annually, the available grid capacity for them could more than double. 

Understanding why requires distinguishing between different workload types. Approximately 80% to 90% of AI computing energy is devoted to inference tasks, with the remaining 10% to 20% consumed by model training. Real-time inference tasks are extremely latency-sensitive and cannot be shifted, but batch inference, offline scheduled inference, and certain training workloads are substantially more flexible and can be paused, queued, or shifted to off-peak or high-renewable periods without affecting user-facing services. 

International pilots have demonstrated this at commercial scale. The EPRI DCFlex initiative, launched in October 2024 with 40 members including Google, Meta, NVIDIA, ERCOT, and NYPA, demonstrated that AI compute clusters can reduce power usage by 25% for 3 hours during peak grid demand. Oracle’s EmeraldAI implementation in Arizona achieved the same reduction through workload tiering, and Google’s partnership with Taiwan Power Company in 2022-2023 demonstrated the applicability of this approach to Asian grid contexts where supply margins are constrained. None of these demonstrations required interrupting critical services.

Malaysia can leapfrog to the strategic planning phase

Malaysia has moved quickly to bring order to a chaotic investment surge. Electricity tariffs have been restructured so that data centers pay costs that reflect the actual burden they place on the system; data centers must now commit to consuming at least 85% of the power they reserve; connection timelines have been compressed from 40 months to as little as 12 months; and the Corporate Renewable Energy Supply Scheme (CRESS) allows large consumers to contract directly with renewable energy producers. What these reforms lack is proper treatment of data center load as something the grid can actively work with. In current planning models that load is still assumed to be fixed, constant, and impossible to shift.

Several features of Malaysia’s market make it a credible place to attempt this shift. The facilities being built are predominantly Tier 3 and Tier 4 data centers equipped with sophisticated monitoring and control systems, modular power infrastructure, and on-site storage. And because most data centers draw power directly from the national grid rather than running on private generation, they remain connected to the utility and system operator in ways that make regulatory engagement possible. What is missing is the analytical framework to value flexibility in planning models and the regulatory architecture to transact it.

A Path Forward

Moving from a supply-centric to system-optimization-centric for data center energy planning requires:

1. Making data center demand response a credible planning resource: Regulatory mechanisms must formally recognize demand-side flexibility as substitutable for a portion of firm generation capacity in adequacy planning. Ancillary services market products need to be developed so that data centers can offer fast frequency response and operating reserves through load reduction. Real-time data exchange must be institutionalized through standardized, auditable data streams embedded into operational and planning processes.
2. Prioritizing renewable dispatch through verified demand-side flexibility: Gas-fired generation currently serves as the default balancing and ramping resource because data center load is treated as inflexible and solar variability must therefore be absorbed by thermal capacity rather than demand modulation. When flexible loads are made visible to the system operator and incorporated into reserve planning, gas plants can move toward a more limited residual flexibility role, reducing curtailment during midday solar peaks from Malaysia’s Large Scale Solar program and lowering the need to keep gas units online purely for balancing purposes.

Regional Implications

As the fastest-growing data center market in Southeast Asia, the energy planning precedents Malaysia sets will be observed and adapted across a region where the stakes are high. An estimated $45 to $75 billion in solar and wind investment will be needed across the region’s top five data center markets by 2030, and without more strategic planning, the default trajectory for most of these markets is a modest renewable PPA laid on top of continued fossil fuel expansion, with grid financing that struggles to keep pace with demand growth. Thailand offers a partial counterexample: its Power Development Plan 2024 embeds demand response and smart grid targets including 2,000 MW in peak reduction measures, but its data center incentive framework rewards power usage efficiency without requiring active grid flexibility participation, leaving demand response ambitions and rapid buildout advancing separately rather than as an integrated system. 

Within the emerging architecture of low carbon ASEAN Power Grid, these dynamics become regionally coupled: large digital loads can either entrench fossil-based marginal supply or anchor cross-border renewable optimisation. Malaysia’s planning framework thus functions as a de facto input into APG design, shaping whether integration locks in existing dispatch patterns or enables a lower-carbon regional equilibrium.  

With both institutional momentum and market scale, Malaysia is well positioned to demonstrate  that large loads (from data centres to other high-density electrified demand centres) can actively reduce system costs and enable deeper renewable integration rather than simply triggering more supply procurement. The frameworks Malaysia develops will shape how the rest of Southeast Asia responds to the same pressures, and the energy transition implications will be determined not by the scale of investment but by the planning frameworks that shape it.