Malaysia’s push to become a regional data centre hub is putting pressure on its energy infrastructure, as new power generation and grid capacity take much longer to build than the facilities they are meant to support, according to experts.

This gap could widen as data centres take up a growing share of electricity demand, which means Malaysia needs to align its investment approvals with its plans for power generation, grid capacity and renewable energy, says Asian Strategy and Leadership Institute (ASLI) head of advisory and research advocacy Dr Fiaz Ahmad.

The challenge, he says, is making sure the infrastructure and policies behind these investments develop at the same pace as the projects themselves.

“Malaysia is not short of policy; it is short of orchestration. We have to see the connection between the opportunities and the demands that we are creating and that these data centres are bringing,” he says during a panel discussion on artificial intelligence (AI), data centres and the energy transition at CelcomDigi Responsible Business Day 2026, held at the telco’s headquarters in Petaling Jaya on Nov 5.

The sector has already drawn large sums, with Malaysia recording RM385.7 billion in data centre-related investment approvals between 2021 and the first half of 2026. Its direct contribution to gross domestic product also grew from RM0.9 billion in 2021 to RM14.1 billion in 2025, according to Bank Negara Malaysia estimates.

Fiaz says data centres are expected to make up about 21% of Malaysia’s electricity demand by 2030 and 31% by 2035, up from about 7% today. Much of this demand has yet to materialise, however, as only one of the 27 projects approved by the data centres task force is operational, he notes.

“Most of these costs and most of the benefits are still ahead of us,” he says, adding that Malaysia must plan for the infrastructure and energy needs alongside the economic opportunities, rather than treat these needs as reasons to reject data centre investments.

The other panellist, Ricardo Reina, founder of Singapore-based energy transition advisory firm Catalyze Transition, says the problem is made worse because AI infrastructure and the energy systems needed to support it are built on very different timelines.

In his presentation, Reina breaks AI infrastructure down into five layers, namely applications, AI models, computing hardware, data centre facilities and the energy system. Applications and models can be developed within months, while computing infrastructure such as graphics processing units (GPUs), servers and networking can take one to two years to set up. Data centre facilities usually take two to three years, but the energy system needs three to seven years.

“Whereas many of these data centres and facilities are being built, developed and conceptualised in a matter of two to three years, the energy system needs three to seven, even 10 years in some places, from planning to execution,” he says.

He notes that much of the grid capacity coming online today was planned in 2016, before the current wave of AI development and the resulting demand for data centres became clear.

The mismatch is most visible in the grid, Reina says, as utility companies must plan and build infrastructure years before it can be switched on, while data centre operators want to be connected much sooner. In his view, the main constraint lies in the grid rather than in the supply of power itself.

He also points out that data centres consume electricity at a steady, high level regardless of the time of day. This makes it harder to meet their demand with renewable energy, particularly solar power, which depends on sunlight.

He adds that hyperscalers, the largest cloud computing providers, are increasingly seeking round-the-clock renewable energy supply, not only to meet their sustainability commitments but also to manage their electricity costs.

“Many of them are aiming for 24/7 renewables, which means that every single hour of the day, I want to be powered by 100% renewables. Many of them, yes, they are willing to consume solar power during the day, but they also want to have visibility on what are going to be their costs for power at night, in the early morning when the sun is not shining,” he says.

Securing round-the-clock supply, he says, allows operators to guard against swings in electricity prices when solar power is not available.

Reina says Malaysia has a relatively advanced regulatory framework for renewable energy compared with other countries in the region, but needs to be more ambitious in adding capacity.

Malaysia has historically added about 1GW of solar capacity a year, a rate Reina believes can be raised to meet future demand. The pace of renewable energy growth remains a concern, however, as most of the recent increase in electricity generation has come from fossil fuels.

According to Reina’s presentation, Malaysia’s electricity generation rose by 7.9 terawatt-hours (TWh), or 4.3%, in the 12 months to May 2026. Fossil fuels made up 92% of that increase, adding 7.3TWh, while hydro, solar and bioenergy together added just 0.6TWh.

Natural gas contributed the most at 5TWh, followed by other fossil fuels at 1.4TWh and coal at 0.9TWh.

On the policy front, Fiaz says the government has introduced a requirement for data centres to source at least 30% of their electricity from renewable sources, which creates more demand for solar generation and battery storage.

Citing estimates, he says applying the requirement to new projects could create demand for about 9GW of solar capacity and 4.5GW of battery storage, a market potentially worth RM36 billion.

Whether that demand turns into actual investment, however, depends on how quickly renewable energy projects can be built and connected to the grid.

The discussion also turns to who should pay for this expansion in renewable energy supply, with Reina arguing that large technology companies, particularly hyperscalers, can afford to pay a premium for reliable and cleaner electricity rather than rely entirely on existing infrastructure.

“They will not go out and say, ‘Hey, just charge me.’ But the truth is it’s much more expensive for them to not be able to build a lot of this infrastructure, or to have a data centre built and waiting for a connection, than to pay a premium for the energy or the green technology,” he says.

He says some companies are already taking a more direct role in securing their energy supply by building substations and power plants on site, while also taking care to keep public support by not passing extra costs on to other electricity consumers.

Looking abroad, Fiaz points to Singapore’s approach of allocating data centre capacity based on sustainability considerations, as well as Ireland’s efforts to link new data centre developments with additional power generation.

He also cites Google’s recent investment in water infrastructure in Malaysia as an example of how large investors can help build facilities that serve both industry and local communities.

However, he cautions that having policies and targets is not enough if there is no accurate way to measure whether they are being implemented.

His own research, which tracks 1,000 Malaysian-listed companies from 2018 to 2023, finds that only about 300 regularly report their carbon emissions, even though sustainability reporting is compulsory for publicly listed companies.

He argues that the same principle should apply to data centre investments and the energy infrastructure behind them, with progress judged by measurable results rather than by approvals and plans alone.