Data centre electricity demand is set to quintuple by 2034 and the grid plan already carries the number

The electricity supply business plan covering 2025 to 2034, known as the RUPTL, carries a projection for data centre demand that has attracted less attention than it deserves. Demand is put at roughly 1,098 megawatts in 2025, about 2,122 megawatts in 2026, and around 5,226 megawatts by 2034, of which 4,621 megawatts would sit within PLN's main system.

The near doubling between 2025 and 2026 is the part worth pausing on. A load that roughly doubles within a single year behaves less like ordinary demand growth and more like the arrival of a new industrial sector. For comparison, electricity consumption in Bali grew 8.02 percent in the first half of 2026 and that was among the highest rates in Indonesia.

What is driving it

Gregorius Adi Trianto, Executive Vice President at PLN, identified artificial intelligence as a significant driver alongside cloud computing and digital services. The distinction matters technically. An AI workload is not simply a larger version of conventional hosting. Graphics processing units draw far more power per rack, which raises power density and imposes heavier cooling requirements on the same floor area.

The geography is also concentrated. Existing capacity clusters around Cikarang, Karawang and the wider Jakarta region, drawing on the Java and Bali system. That concentration means the load growth lands on specific substations and specific transmission corridors rather than spreading evenly across the national system.

Individual commitments give a sense of the scale involved. PLN and BDx Data Centers, operating through PT Starone Mitra Telekomunikasi, agreed arrangements covering a combined 1.2 gigawatts. The first tranche is 788 MVA for the CGK4 facility at Jatiluhur in West Java, delivered in three phases, with 60 MVA for CGK3A in South Jakarta and 385 MVA for CGK5 in the Suryacipta industrial area. PLN is building a 150 kilovolt substation at Jatiluhur to serve the zone.

The supply side is already dimensioned for it

The encouraging feature of this picture is that the demand projection and the supply plan sit in the same document.

PLN plans to add 69.5 gigawatts of generation and storage capacity through 2034, with renewables and battery storage accounting for roughly 76 percent of the total. Against that, 5,226 megawatts of data centre load in 2034 is a demanding but accommodated number rather than a surprise.

A load that doubles in a year is only a crisis if nobody wrote it down in advance. This one is in the plan, with the capacity to serve it in the same plan.

Hendra Suryakusuma, chairman of the Indonesian data centre association IDPRO, identified the real constraint precisely. Adding generation is not sufficient on its own. Transmission networks, substations and access to renewable supply have to develop at the same time, and they have to do so against data centre construction timelines that are considerably shorter than grid infrastructure timelines.

That mismatch in build times, rather than any shortage of planned megawatts, is where the difficulty actually sits. A data centre hall can be built and commissioned in well under two years. A transmission line or a new substation, from route selection through land acquisition to energisation, routinely takes longer. When the two are started at the same moment, the building finishes first and waits.

The practical implication is that the useful unit of planning is not the national megawatt total but the specific substation. A national reserve margin that looks comfortable in aggregate tells a planner very little about whether the Cikarang corridor can accept another 200 megawatts next year. That question is answered at the level of individual network assets, and it is where the attention belongs.

Why this belongs in an energy efficiency discussion

Data centres are usually framed as a demand problem. They are also one of the few large loads where efficiency is measured continuously, reported publicly and treated as a competitive variable.

The industry tracks power usage effectiveness, the ratio of total facility energy to the energy reaching the computing equipment itself. Everything above that ratio is overhead, most of it cooling. In a tropical climate that overhead is structurally higher than in temperate locations, which makes cooling design the single largest efficiency lever available to an Indonesian operator.

This is also a load with unusual flexibility in where it sits. Unlike a factory tied to a port or a mine tied to a deposit, computing capacity can in principle be placed wherever power and connectivity allow. That gives planners a lever that most industrial siting decisions do not offer.

There is a further property worth noting. A data centre runs at a high and steady load around the clock, which makes it close to an ideal customer for generation that cannot easily be throttled, and a demanding one for solar without storage. That steadiness is why operators pursuing renewable supply tend to contract for geothermal, hydro or firmed solar rather than relying on daylight generation alone, and it is one reason the geothermal capacity being recovered at existing fields has a ready market.

Sirkularium's read for government and public institutions

Three observations follow.

The first concerns using siting as an instrument. If data centre load is concentrating on the Java and Bali system while the 100 gigawatt peak solar programme is building capacity across six provinces, there is an argument for encouraging some of that load toward where new generation is arriving. Connectivity requirements constrain how far this can go, but incentives that shape location are cheaper than transmission built to follow load that chose its own address.

The second concerns efficiency standards for a sector that already measures itself. Because power usage effectiveness is reported as a matter of commercial practice, a reporting requirement imposes very little additional burden. Making disclosure standard for facilities above a threshold size would give planners visibility into how much of this growing load is computing and how much is cooling overhead, which is exactly the information needed to judge where efficiency policy would pay.

The third concerns matching the clean supply to the load. Data centre operators serving international clients increasingly face customer requirements for renewable supply, which turns clean generation into a commercial advantage rather than an obligation. Aligning the renewable share of the 69.5 gigawatt programme with the locations where data centres are being built would serve the grid and the export services sector at the same time.

What to watch next is whether transmission and substation delivery keeps pace with facility construction, whether data centre siting broadens beyond the Jakarta corridor, and whether the 2026 demand figure of 2,122 megawatts is met, exceeded or undershot when the year closes.