CAAICE
25 July 2026

The constraint moved

Johor is rejecting roughly a third of data centre applications while Malaysia contracts gigawatts of new supply. The two facts are not in tension. They describe a change in what these projects are actually competing for.

Two things have been reported about Malaysian data centres over the past eighteen months, and read together they look contradictory.

Johor established a coordination committee to vet data centre applications in June 2024. Roughly thirty per cent of applications have been turned down. In March the Menteri Besar said the state would not approve data centres with high water demand, and that public welfare came first. Johor has stopped approving Tier 1 and Tier 2 facilities. At the federal level the Prime Minister has said that proposals for non-AI data centres have been rejected for close to two years.

Over the same period Tenaga Nasional Berhad signed agreements covering forty-nine data centre projects representing as much as 7 GW of potential demand. DayOne contracted for the output of 1.5 GW of solar and 2.2 GWh of storage. Announced capacity under the Corporate Renewable Energy Supply Scheme reached roughly 4 GW.

A state that spent years making itself the most attractive data centre destination in the region is now refusing a third of the business, at the same moment the national utility is signing the largest load agreements in its history.

Both things are true. Understanding why requires looking at what the rejections were actually for.

The stated grounds

The reporting on the rejected applications is reasonably specific. Applications failed on sustainability standards. They failed because operators had chosen sites without adequate utility infrastructure. They failed because utility timelines were misaligned with the developer’s schedule, and because planning problems surfaced late.

Separately, electricity and water approvals have been running to as much as eighteen months.

None of these are objections to data centres. A state that wanted to discourage the sector would raise the land price, tighten the zoning, or simply stop returning calls. Johor did none of that. It set up a committee and started asking whether the applicant had thought about where the power and water were going to come from.

That is not a policy reversal. It is a sequencing failure being caught at the door.

Two clocks

The bodies that attract investment and the bodies that allocate utilities are not the same institution, do not report to the same principal, and are not measured on the same thing.

An investment promotion agency is judged on committed capital and jobs announced. It moves fast because moving fast is the product. Land, incentives, approvals in principle: months.

A utility is judged on system reliability and cost recovery. It moves at the speed of substation construction and generation adequacy planning. Water authorities move slower still, because water supply is a public health function before it is an industrial input, and because the political consequences of a shortage are immediate and personal in a way that a delayed connection is not.

A developer can therefore complete the first process well before it has meaningfully begun the second. Land is secured. Incentives are agreed. Board approval is obtained on a timeline built around the fast clock. Then the connection application enters a queue that was never going to clear inside the assumed schedule, and the project discovers that the thing it treated as an administrative formality is the actual constraint.

The rejection is the moment the two clocks are reconciled. From the outside it reads as a state changing its mind. From inside the process it is the second institution declining to absorb an assumption it was never asked about.

What this means for how these deals are built

The practical consequence is that the binding constraint on a Southeast Asian data centre project has moved, and transaction practice has not caught up with it.

For twenty years the constraints on this kind of industrial siting were land, tax treatment and construction permitting. Diligence was built to test those. Conditions precedent were drafted around them. Long-stop dates were set by reference to how long it takes to get planning consent and build.

The constraint is now utility allocation. Grid connection capacity at the specific point of interconnection. Water allocation from a specific catchment, in a state where the political economy of water is contested. Neither is fungible, neither can be bought at a premium, and neither can be accelerated by hiring better advisers.

Some observations follow from that, and they are not complicated:

Grid connection status belongs in first-round diligence, not in the technical annexes. The question is not whether capacity exists in the state but whether it exists at the substation serving the site, and where the applicant sits in the queue behind everyone else who asked first.

Long-stop dates set by reference to construction schedules are the wrong instrument when the critical path runs through a utility queue with an eighteen-month tail and no contractual commitment to a date. A long-stop that expires before the connection could plausibly issue is a termination right dressed up as a deadline.

Conditions precedent should distinguish between an approval in principle and an allocation. These are frequently conflated in term sheets, and they are not the same asset. One is a statement of policy support. The other is a claim on a scarce physical resource.

Water is now a separate diligence workstream with its own regulatory counterparty, its own timeline and its own political risk. Treating it as a subheading under environmental permitting understates what has changed.

None of this is exotic. It is ordinary infrastructure practice applied to an asset class that, until recently, did not need it. Data centres were commercial real estate with servers in them. They are now, functionally, industrial facilities with utility-scale load, and they are being procured by teams whose instincts were formed on the earlier model.

The tier criterion is doing something else

One element of Johor’s approach deserves closer attention than it has received.

The state has said it will stop approving Tier 1 and Tier 2 facilities, on the stated basis that they consume something in the order of two hundred times the water of Tier 3 and Tier 4 facilities.

The tier classification, as it is ordinarily used, describes redundancy and concurrent maintainability. It is a measure of how much duplicated infrastructure a facility carries so that it can stay up during a component failure. It is not a measure of cooling technology, and it does not on its face predict water intensity. If anything the intuition runs the other way, since higher-tier facilities carry more of everything.

Which suggests the tier criterion is functioning as a proxy for something the state has not articulated directly. Plausible candidates: operator sophistication, capital commitment, cooling system design, or simply the size of the counterparty and the likelihood that it will engage seriously with efficiency conditions.

Proxies are not illegitimate. Regulators use them constantly, because the thing they actually care about is hard to measure and the proxy is easy to verify. But a proxy that does not track the underlying variable will produce both false positives and false negatives, and applicants will eventually work out how to satisfy the proxy without satisfying the objective.

If the real concern is water intensity, the criterion should be water intensity, expressed as a number, with a measurement method. That would be harder to draft and much harder to argue with.

What travels

Johor is ahead of its neighbours in this, not behind them. It attracted the load first and hit the constraint first.

Every jurisdiction in the region that succeeds in attracting this investment will arrive at the same place, and most of them have the same institutional structure: a fast, well-resourced investment promotion function sitting alongside a slower utility with its own mandate and no obligation to underwrite the promises made upstream.

The interesting question for the next two years is not which country attracts the most data centre capital. It is which one first builds a process where the utility is in the room at the point the investment is being solicited, rather than eighteen months later when the application arrives and the answer is no.

CAAICE publishes all research free of charge. This brief draws on publicly available sources and does not constitute legal advice. Corrections and challenges are welcome at [email protected].

Sources