Key points

Data center growth creates opportunities for Indonesia’s digital economy. Computing capacity also needs electricity, cooling water, and governance that protects nearby communities. Discussing AI means discussing the physical infrastructure behind it.

IndicatorFigure reported in the sourcesHow to read it
National data center power pipeline25,297 MW by 2034A pipeline projection, not today’s operational capacity.1
Planned BDx Jatiluhur capacity640 MWThe project’s planned capacity.2
BDx Jatiluhur water requirementsUp to 384 liters/second, about 33,000 m³/dayProjected demand at full capacity, not measured current consumption.2
Batam project water requirementsAbout 3 million liters/day at Kabil and 29 million liters/day at NongsaRequirements for new and proposed facilities, not evidence that all are operating.3
Illustrative energy demand at 640 MWAbout 5.6 TWh/yearA calculation assuming a continuous 640 MW load for 8,760 hours, not facility meter data.

Digital economic ambitions and resource limits

Digital transformation and artificial intelligence are being promoted as drivers of national economic growth. Yet discussions of the cloud and AI often seem detached from their physical requirements: buildings, servers, power networks, cooling systems, and water.

S&P Global’s report on government and PLN presentations at Enlit Asia 2026 describes an RPJMN 2025–2029 data center capacity target rising from 0.74 watts per capita in 2024 to 6.87 watts per capita in 2029. It also reports a power pipeline of up to 25,297 MW by 2034, alongside approximately 1,990.5 MVA of connected capacity at the time.1

MW and MVA are not interchangeable. MW measures active power; MVA measures apparent power. These figures need to be read with their definitions, reporting dates, and project status, rather than presented as a direct comparison of identical measures.

Development patterns: concentration and two local cases

Concentration on the Java–Madura–Bali grid

According to that report, the Java–Madura–Bali grid served 142 of PLN’s 159 data center customers in July 2026 and accounted for approximately 99% of the reported connected capacity. Pipeline additions are also concentrated, with projections of 6,884 MW in 2027 and 7,798 MW in 2028.1

An electricity customer does not necessarily correspond to one building or one company. A project pipeline also does not guarantee that every project will be completed. Nevertheless, concentrated demand is a reason to plan generation, transmission, substations, and water resources together.

BDx Jatiluhur: permits and transparency before construction

In October 2026, ANTARA reported that the West Java Provincial Government had halted construction of BDx’s planned 640 MW AI data center in Jatiluhur, Purwakarta. The report described issues involving environmental impact assessment and building approval documents, alongside projected water requirements of up to 384 liters per second at full capacity.2

IESR supported the temporary halt and called for public disclosure of studies covering environmental impacts, water requirements, electricity reliability, and effects on PLN customers. The questions include who receives access to resources and who bears the costs and risks.4

INTIMEDIA also reported WALHI’s estimate that the project’s water allocation could equal the daily clean-water needs of more than 220,000 people. This comparison is the organization’s estimate, not evidence that community supplies have already been diverted or that the impacts have been measured.5

Batam: projected demand and residents’ experiences

Global Voices reported requirements of around 56 MW and up to 3 million liters of water per day for new facilities at Kabil. Nine proposed projects at Nongsa were reported to need around 285 MW and 29 million liters of water per day. The report estimated that existing and planned facilities could use about eight percent of Batam’s water supply.3

It also documented residents’ protests in September and December 2024 over water supply disruptions. Authorities attributed the December distribution problem to a technical fault. These protests and disruptions should therefore be recorded without assuming that data center operations caused every water problem.3

Environmental impacts that need measurement

Water: withdrawal, consumption, and location

Evaporative cooling can reduce energy demand while consuming water through evaporation. Its impact depends on cooling technology, climate, water sources, and watershed conditions. Good electricity efficiency does not necessarily mean good water efficiency.6

Li and colleagues also distinguish direct water use from the indirect water footprint associated with electricity supply. Assessments of AI infrastructure need clear accounting boundaries so that comparisons measure the same things.7

Electricity: planned capacity is not actual energy consumption

For illustration, a constant 640 MW load operating for a full year would require approximately 5.6 TWh: 640 MW × 8,760 hours = 5,606,400 MWh. This is a mathematical scenario, not a report of BDx consumption. Actual use depends on utilization, commissioning stages, and whether the capacity figure refers to IT load or the entire facility.

The electricity source also determines the emissions impact. New computing capacity needs to be planned alongside clean energy supply, rather than assumed sustainable simply because the resulting services are digital.1

Governance: permits need public evidence behind them

MAP UGM’s discussion of digital decarbonization connects technology growth with the national net-zero agenda. The questions include alignment between digital and climate policy, energy measurement, and accountability for infrastructure impacts.8

Technical standards or permits alone do not explain the cumulative impact of several facilities in one area. Communities need evidence they can examine: water sources, seasonal requirements, electricity consumption, emissions, and plans for droughts or supply constraints.

Policy directions and technical choices

Moratorium proposals and community protection

WALHI, as reported by INTIMEDIA, has called for a moratorium on new construction while dedicated AI infrastructure safeguards are developed. Its proposals address environmental carrying capacity, transparency, and participation by affected communities. This is an advocacy position, not a statement that a nationwide moratorium is already in force.5

Efficiency that can be audited

In its statement, IESR proposed waterless or closed-loop cooling, a PUE limit below 1.3, and direct renewable energy use of at least 35% by 2030 and 50% by 2040. These are IESR recommendations, not national obligations described as already in force.4

Cooling choices need to be assessed as complete systems. A closed loop on the server side does not automatically eliminate water use on the heat-rejection side. Evaluation should include make-up water, energy, wastewater treatment, and dry-season requirements.6

Planning energy and water from the start

Reporting on Danantara’s position for projects above 500 MW emphasizes independent energy provision alongside water reuse and treatment. Independent generation still needs to be assessed by its source and emissions; a dedicated power plant is not automatically clean energy.9

Desalination and reclaimed water also require assessment of energy demand, location, costs, and environmental impacts. The aim is to ensure that computing requirements do not reduce the reliability of basic services for households, agriculture, and other industries.

Closing thoughts

Digital progress needs to be evaluated alongside its physical costs. For Indonesia, measures of data center success should include economic benefits, resource efficiency, transparency, and protection for communities.

People working in technology can start with concrete questions: what need does this computing serve, how many resources does it require, where do those resources come from, and who bears the impacts?

References

  1. Rachel Tan, S&P Global Energy (22 September 2026). Indonesia maps grid strategy for hyperscale data center boom. Reporting on government and PLN presentations; pipeline figures are projections. ↩ ↩2 ↩3 ↩4

  2. ANTARA News (12 October 2026). Minister AHY urges green compliance as W. Java halts AI data center. ↩ ↩2 ↩3

  3. Mong Palatino, Global Voices (21 April 2026). AI-ready data centers are booming in Indonesia but water woes loom. ↩ ↩2 ↩3

  4. IESR (1 October 2026). Statement on the temporary BDx Jatiluhur halt and proposed AI safeguards. ↩ ↩2

  5. INTIMEDIA (October 2026). AI Data Center Expansion Threatens Community Water Resources, Government Urged to Impose Construction Moratorium. Reporting WALHI’s estimates and proposals. ↩ ↩2

  6. Patrick J. Campbell, Ramboll (2 December 2025). How to use water wisely in data centres. ↩ ↩2

  7. Pengfei Li, Jianyi Yang, Mohammad A. Islam, and Shaolei Ren (2023; revised 2025). Making AI Less “Thirsty”: Uncovering and Addressing the Secret Water Footprint of AI Models. ↩

  8. MAP UGM, Buletin INSIGHT 8(3). Digital Decarbonization & Indonesia’s Net-Zero Agenda. ↩

  9. CNN Indonesia (23 October 2026). Danantara Minta Proyek Data Center AI Bangun Sumber Energi Sendiri. Available syndicated summary. ↩