The Land Is Cheap. The Power Is Not: Asia's Data-Centre Squeeze
Hyperscale demand has turned obscure industrial plots into the most contested real estate in Asia. But the deals are increasingly won and lost on the electricity grid, not the title deed.
- ·Capital rotates out of US/EU equities into hard ASEAN infrastructure.
- ·Data centres, power transmission and ports are the three priority lanes.
- ·Vietnam, Indonesia and the Philippines absorb the largest allocations.
A decade ago, a 20-hectare industrial plot an hour outside Kuala Lumpur in the corridor toward Johor would have been valued mostly on its proximity to a port and a highway. Today it is valued on something invisible: how many megawatts the local utility can deliver to its fence line, and how quickly. That single question now moves prices more than location, zoning or soil ever did.
The shift is the clearest sign yet of how the artificial-intelligence boom has rewired Asian real estate. Hyperscale data centres, the warehouse-scale buildings that train and serve AI models, have become the region's hungriest new tenant. They want flat land, fibre, water for cooling and, above all, electricity at a scale that strains national grids built for a slower era.
Land brokers across Malaysia, Indonesia, India and Japan describe the same scramble: pension funds, infrastructure giants and cloud companies racing to lock up sites, often years before they intend to build. The result is a market where the dirt is plentiful and comparatively cheap, but the power connection has become the scarcest, most valuable thing in the deal.
The grid is the bottleneck
Talk to anyone closing data-centre land deals in 2026 and the conversation turns within minutes to grid capacity. A single large campus can draw 100 megawatts or more, enough to power a small city. Utilities that once approved such loads in weeks now quote connection timelines stretching to 2029 and beyond, particularly around the saturated hubs where the first wave of campuses already cluster.
That scarcity has spawned a peculiar secondary market: not in land, but in grid allocations. Developers who secured power-reservation agreements early are finding those rights are worth more than the plots they sit on. In several cases reviewed by BriefAsia, a site with a firm 80-megawatt connection commanded a premium of more than double an identical neighbouring plot with no allocation, even though the buildings on top would be indistinguishable.
The bottleneck is reshaping geography. Demand is spilling out of crowded primary markets into second-tier locations chosen almost entirely for spare grid headroom, places with a large idle substation or a decommissioned heavy-industry site whose old power feed can be repurposed. An aluminium smelter that closed years ago is suddenly prime data-centre land, not for its buildings but for the fat cables still running to its gate.
We used to underwrite these deals on land, fibre and tax. Now the first slide is always the grid. If the electrons cannot get there by 2028, nothing else on the page matters, said a Singapore-based infrastructure fund partner who asked not to be named.
Water, heat and the new permits
Power is the headline constraint, but it is not the only one. Cooling a hyperscale facility consumes enormous quantities of water or, increasingly, requires expensive closed-loop and liquid-cooling systems to avoid it. In water-stressed parts of India and northern China, regulators have begun treating a campus's projected water draw as a gating permit question, separate from its energy plan.
Heat is becoming a planning variable too. Cities that once welcomed any industrial investment are starting to ask harder questions about what a cluster of always-on campuses does to a neighbourhood and a grid during a heatwave. A handful of municipalities now require waste-heat reuse plans, pushing developers to pipe excess warmth into district heating or nearby greenhouses rather than simply venting it.
These layered requirements have lengthened the approval gauntlet. A campus that might have moved from option to permit in a year now routinely takes two or three, as energy, water and environmental reviews stack up. For investors underwriting eight-figure land positions, that timeline is itself a risk, and one increasingly priced into the deal.
Who is buying, and why now
The capital chasing these sites is unusually patient and unusually large. Sovereign and pension money, drawn by long-dated, inflation-linked income from contracted hyperscale tenants, has poured into platforms assembling land banks across the region. The thesis is simple: AI compute demand will keep rising, the buildings to serve it must sit somewhere, and the entities that control the scarce power-ready sites will collect rent for decades.
That thesis is not risk-free. The same logic that makes a power-ready plot valuable makes the market vulnerable to a single shift in the AI economy, a more efficient generation of chips, a slowdown in model training, or a regulatory cap on energy-intensive computing. Land bought today at grid-scarcity prices assumes that scarcity persists. If utilities catch up faster than expected, the premium evaporates.
For now, the imbalance is widening, not narrowing. National grid operators in several Asian markets have published capacity-expansion plans that, even if delivered on schedule, fall short of the connection requests already queued. That mismatch is the clearest signal that, whatever happens to AI fashion, the contest for power-ready land has further to run.
The strategic stakes
Beneath the deal-making sits a larger question of national strategy. Governments increasingly see data-centre capacity as economic infrastructure, a prerequisite for hosting the AI services their companies and citizens will depend on. Several are now bundling power, land and fibre into pre-packaged zones, effectively competing to be the region's compute hub, much as they once competed for semiconductor fabs and electronics assembly.
That competition cuts both ways for the property market. State-assembled zones can short-circuit the grid bottleneck, unlocking sites that private developers could not power on their own. But they can also flood specific corridors with supply, compressing the very land premiums that drew private capital in the first place. The smartest investors are now reading energy ministry roadmaps as closely as any land registry.
The land, in the end, was never the hard part. Asia has plenty of flat, well-connected dirt. What it does not yet have, in anything like sufficient quantity, is the power to animate it. Until the grids catch up, the most valuable real estate in the region will keep being measured not in square metres, but in megawatts.