The data-centre energy conversation tends to fixate on two things: the chips that draw the power and the chillers that cool them. The building itself \u2014 the shell wrapped around all that equipment \u2014 barely rates a mention. In a temperate climate, that omission is forgivable. In tropical ASEAN, it is a measurable cost.<\/p>\n
In a data centre, cooling is not a side cost \u2014 it is a large fraction of the whole. The IEA notes that cooling systems range from around 7% of total consumption in efficient hyperscale facilities to over 30% in less efficient ones<\/strong>. That spread is the entire game: the difference between a competitive facility and an expensive one is largely the difference in how much energy is spent moving heat out.<\/p>\n And the demand is exploding. Wood Mackenzie projects Southeast Asian data-centre power demand to quadruple from 2.6 GW to 10.7 GW between 2025 and 2035<\/strong>. In Singapore, data centres already account for roughly 9% of national electricity<\/strong>. At that scale, a few percentage points of avoidable cooling load is a national-grid-level number \u2014 and a serious line on the operator\u2019s P&L.<\/p>\n Operators pour enormous effort into mechanical and electrical efficiency \u2014 chiller plant, airflow management, the relentless pursuit of a lower PUE. Far less attention goes to the passive shell, yet in equatorial heat the envelope works against the cooling system every hour:<\/p>\n None of this shows up in a server-room dashboard. It shows up as a cooling system that runs harder than the IT load alone would explain \u2014 a PUE that is stubbornly higher than the mechanical design promised.<\/p>\n Improving PUE through mechanical upgrades is capital-intensive and complex. Sealing the envelope and fixing its thermal weaknesses is comparatively cheap, and it reduces the load before<\/em> it ever reaches the chillers. Every watt of solar gain or infiltration eliminated at the shell is a watt the cooling plant never has to fight \u2014 a permanent reduction that compounds across the life of the facility.<\/p>\n The obstacle is the same one that affects every large building: the envelope is rarely measured. A drone thermal survey reads the entire shell \u2014 roof, walls, and junctions \u2014 and locates exactly where heat is entering and conditioned air is escaping. For an operator running multiple facilities across the region, it benchmarks each building\u2019s envelope performance on one scale, pointing remediation at the sites adding the most avoidable load.<\/p>\n As ASEAN races to build the data centres the AI era demands, the operators who treat the building shell as part of the cooling strategy \u2014 not just the box it sits in \u2014 will run the most competitive PUE in the hardest climate to do it. The envelope is the overlooked variable. It is also one of the cheapest to fix.<\/p>\nWhere the envelope enters the equation<\/h2>\n
\n
The envelope is the cheapest part of the PUE problem<\/h2>\n
Measure the shell, not just the racks<\/h2>\n