Where do these chips, manufactured in Taiwan and Korea, ultimately reside? Increasingly, they are deployed in Southeast Asian data centers. This scribble examines the physical constraints emerging in Malaysia, the region’s new “AI Factory.”
The 2.53 GW Expansion
Malaysia’s data center capacity is forecast to double from 1.26 GW to 2.53 GW between 2025 and 2030. This explosive growth is concentrated in Johor, serving as a spillover hub for Singapore. The sector is expected to generate $34 billion in economic output.
The Energy Tariff Shock
However, the laws of physics and economics are asserting themselves. The density of AI workloads (approaching 100 kW per rack) places immense strain on the grid.
-
Tariff Hike: To manage this, Tenaga Nasional Berhad (TNB) has raised base electricity tariffs for the 2025-2027 Regulatory Period (RP4) to 45.62 sen/kWh, a 14.2% increase.
-
Voltage-Based Pricing: The new tariff structure penalizes ultra-high voltage users like data centers with higher demand charges.
This marks a pivotal shift. In 2024, Malaysia competed on cheap land and power. In 2026, it competes on availability. The “Green Lane” for data centers is no longer about speed of approval, but access to electrons. The “Corporate Renewable Energy Supply Scheme” (CRESS) allows data centers to source green power directly, but the grid interconnection fees are rising. This establishes energy security as the primary bottleneck for the continued expansion of the Asian AI ecosystem.