The Data Centres Came. Now Malaysia Is Making Sure They Leave Something Behind.

On 24 July, Malaysia’s Treasury secretary-general, Johan Mahmood Merican, admitted something governments in the middle of a boom almost never say out loud: that they had been too good at their job. “We perhaps overdid the red carpet for data centres,” he said — so successfully that demand is now running ahead of what the country wants to give it. So the rules are changing. Instead of rewarding a project for the size of its cheque, Malaysia is building a “scorecard” that asks what the project actually leaves behind. The full version is due around the middle of 2027.

It would be easy to read this as Malaysia cooling on a sector it spent two years courting. The truer word is refining. Having won the investment, the country is now getting deliberate about what it keeps.

It helps to be honest about what a data centre actually delivers. For all the money it takes to build, it employs almost no one. A hyperscale campus — the big ones — usually runs on a hundred to two hundred permanent staff; one site in Texas is contractually bound to just fifty-seven. American studies land on a startling number: about one permanent job for every US$13 million invested, when a job almost anywhere else costs some US$137,000 to create. The building phase is busier — a thousand or more workers for a year or two — but once construction ends, so do most of the jobs.

None of this is an argument against data centres. Where the ground is prepared, the wider gains are real: in the United States, counties that landed their first big facility saw private employment climb four to five per cent over the following years, and information-sector work rise a good deal more. Where it wasn’t, the same money barely moved anything. The lesson is both uncomfortable and useful: a data centre’s benefits are indirect and conditional, and they flow to hosts that have built the means to capture them.

Malaysia has been building exactly that. Around twenty-five subsea cables come ashore on its coasts; the grid is being pushed harder and wider; and in Penang it runs one of the region’s deepest chip assembly-and-testing clusters. The point of the data-centre wave was always to feed that — to climb the technology ladder, not just to store the world’s files. Seen that way, the scorecard is less a gate than a sieve.

And a fairly specific sieve. By the government’s own account it weighs four things, and the size of the investment is not one of them. Does the project make the economy more sophisticated, or does it just park servers on Malaysian soil? Are the jobs skilled and well paid, or a maintenance crew of a few dozen? Does the investor hand over knowledge and skills that stay behind? And does the facility plug into local suppliers and that semiconductor base, or sit apart from it? Under all four is the measure that replaces the old one: the lasting benefit, weighed against the power and land a project consumes.

That weighing is not abstract. On current projections, data centres could be drawing close to a third of Peninsular Malaysia’s electricity by 2035, against roughly seven per cent now. Malaysia has already started charging for it — the current tariff period lifted industrial power rates and tilted them so the heaviest users carry the most. The incentives are the next thing in line for the same treatment.

For an investor, Malaysia is becoming a more selective host. An operator that arrives with jobs, technology or its own power will still be wanted; one that comes only for cheap land and a grid connection will find it a harder sell than a year ago.

And it is not Malaysia’s problem alone. Thailand’s investment applications jumped something like eighty per cent in the first half of this year, most of them chasing the very same AI data centres — and the same appetite for power and land that Malaysia has just started to bill for. Attracting the capital, it turns out, was the easy part. What a country keeps of it is the harder question, and the one now forming quietly across the region.