AI’s power dilemma

FOR years, Malaysia’s challenge was attracting high-value investments.

Today, it faces a different problem: ensuring it has enough electricity to sustain them without passing the cost on to everyone else.

Electricity demand from data centres (DCs) is expected to jump from 10,544 GWh today to 73,274 GWh by 2035, while peak demand in Peninsular Malaysia is forecast to rise by more than 57% over the same period.

Data centres are unlike traditional manufacturing investments. Their appetite for electricity is immense, continuous and only expected to grow further as artificial intelligence (AI) applications become more widespread.

Meeting that demand will require far more than building new power plants.

Transmission lines, substations, reserve capacity and grid modernisation will require sustained capital expenditure running into tens of billions of ringgit over the coming decade.

The government’s decision to extend the life of ageing thermal power plants and accelerate new gas-fired generation illustrates the scale of the challenge.

While necessary in the short term, it also risks locking Malaysia into greater fossil fuel dependence just as it seeks to accelerate its energy transition.

The more pressing question, however, is who ultimately pays for this expansion.

If utilities such as Tenaga Nasional Bhd are required to make significant investments, those costs cannot simply disappear.

They will either be absorbed through regulated returns, supported by government subsidies, or recovered over time through electricity tariffs.

The ideal outcome is one in which DC operators bear a larger share of the infrastructure costs through connection charges, dedicated power facilities and renewable energy (RE) procurement, rather than households and small businesses subsidising the digital economy. The government’s insistence that new DC approvals be aligned with grid capacity is encouraging, as is its push for operators to develop their own RE sources and reduce reliance on the national grid.

But implementation will be crucial.

Screening projects based on available power capacity must remain a genuine exercise, not one that gives way under the pressure to attract investment.

Malaysia also needs to ask whether success should be measured solely by the billions of ringgit in announced investments.

A single DC may consume as much electricity as a small city, yet employ only a fraction of the workforce created by conventional manufacturing.

Malaysia should welcome DCs, but not at the expense of consumers.

It also raises a question that analysts are increasingly debating: whether the economic returns from DCs are commensurate with the enormous demands they place on the national grid.