Malaysia’s economy expanded 5.5% in the first nine months of 2026, according to advance estimates, and a good part of that came from pouring concrete for data centres and shipping chips and servers to buyers abroad. The money is real. What is not yet established is how much of it stays behind as capability once the current wave of construction ends.
Ernst & Young Consulting business consulting partner Nik Anira sets the condition in operational terms. Construction and exports can keep supporting expansion, she says, but sustaining the expansion requires stronger productivity, more domestic value creation and local capabilities that did not exist when the projects were approved. Hiring during construction is the easy part.
Her argument is that a project should support activity after the cranes come down, and that export volume should be judged by whether it strengthens Malaysian firms. For electrical and electronics manufacturers, that means using the country’s semiconductor base to develop products and technologies of their own. The assembly end of someone else’s supply chain is not where that capability accumulates.
“Attracting investment is also important, but we should also consider how it develops local suppliers, supports knowledge transfer and creates higher-value jobs,” she said.
The policy menu she describes is unglamorous: sustained spending on research, development, commercialisation and innovation, closer work between government, industry and universities, and money for prototype development and pilot production, the stage where most research stops being research and starts requiring a factory.
“Success should be measured by stronger Malaysian technological capabilities, greater domestic value creation and higher-value jobs that support higher income levels for Malaysians.” That is a different scorecard from the one used to clear the projects.
There is a second exposure that has little to do with skills. Malaysia’s growth has become tied to AI-related investment and demand, which leaves the economy open to a reversal in the global technology spending cycle.
Carmelo Ferlito, chief executive of the Center for Market Education, describes the construction boom as largely driven by data centres and draws a distinction that matters for how a correction would play out. Most of the money is private, raised or borrowed by investors who carry the loss if demand was misjudged. Buildings financed that way do not become a public liability when the business case weakens.
“These are mostly private investors using their own or borrowed capital, and they will absorb the losses if they’ve misjudged demand. That’s healthy.”
His concern sits in the parts of the boom that rest on electricity, water and land priced below what it costs to supply them, or on incentives that stand in for a business case. A data centre that clears its hurdle rate on a subsidised tariff is not the same asset as one that clears it at cost, and the difference only becomes visible when the tariff is reset.
“Wherever prices don’t reflect true scarcity, investment signals get distorted and we may be overbuilding,” he said.
If AI spending cools and some capacity turns out to have been premature, Ferlito’s position is that the losses should stay with the investors who took the risk. Public money, on this reading, would delay the correction and move its cost to people who took none of the risk.
“That’s the market correcting, and it should be allowed to happen, not offset with public spending,” he said. A rescue would move the cost of a private misjudgement onto people who were never part of the decision.
Lee Heng Guie, executive director of the Socio Economic Research Centre, makes a related point about what this cycle is made of. It leans heavily on data centres and the energy and utility infrastructure built around them, where earlier expansions leaned on government spending on highways, railways and ports. A boom assembled from private capital and external demand behaves differently from one assembled from a public works budget, and it can stop faster. He expects growth of 4.5% to 5% a year from 2027 to 2029, carried by domestic demand and electronics exports, and names a potential AI boom-and-bust cycle as the risk to semiconductor exports.
Ferlito adds that growth concentrated in one investment wave and one category of external demand stays exposed to shifts in the global cycle, however full the current order books look. The composition of that demand matters as much as its size.
Neither point settles the separate argument about public infrastructure, which in Malaysia has its own history and its own politics of delivery. Lee points to the East Coast Rail Link and the Johor Bahru-Singapore Rapid Transit System Link as evidence that large projects remain on the agenda, financed differently. Under what he calls Public-Private Partnerships 3.0, private investors take a larger share of the financing burden and the commercial risk that comes with it. He lists the Penang LRT, the third terminal at Port Klang and the Bandar Malaysia-Seri Kembangan Expressway as examples of a more targeted approach, each tied to a specific bottleneck.
Whether that pipeline can carry the construction industry past the data centre phase is a question nobody in the article answers with a number. The projects named are years from completion, and the employment they create arrives in bursts tied to civil works rather than in a steady line.
Nik Anira argues that Malaysia still needs major infrastructure, especially outside the Klang Valley, to create new growth centres and narrow the gap between the capital region and the rest of the country. Her test is the pipeline of projects still at planning stage.
“I would not say we have necessarily slowed infrastructure spending prematurely simply because many projects currently under way were conceived years ago. Mega projects have long development cycles,” she said.
“The more important question is whether we are building a sufficiently strong pipeline today to meet Malaysia’s needs over the next decade, particularly outside the Klang Valley.”
Fiscal consolidation, in her reading, constrains how projects are selected and financed rather than whether they proceed. That is a narrower claim than it first appears, and it puts the weight on the selection process.
“Fiscal consolidation does not necessarily mean infrastructure investment has to stop. It means being more selective about what we fund, how we fund it and when we deliver it,” she said.
Her priorities run to transport and logistics links, reliable electricity and water, digital connectivity and flood mitigation. Private finance, concessions and public-private partnerships could carry the projects that can pay for themselves, leaving public money for the ones that return value in wider economic and social terms.
The two arguments in the article are compatible, which is what makes the policy question awkward. Private capital absorbing its own losses does not fix an input price that understates scarcity, and disciplined public investment does not close a capability gap in the supplier base. Each addresses a different failure.
What would settle the larger argument is not another headline investment figure. It is evidence that local suppliers are taking a larger share of project spending, that the plants now being built are hiring engineers as well as assemblers, and that the capacity under construction can earn a return once AI hardware demand stops growing at its present rate. None of that is visible in the numbers available today.