Longhand ResearchIndonesia / Infrastructure

The resources behind AI 05 October 2026

AI runs on
power and water.

A server hall is also a claim on the grid and the local water supply. As Indonesia adds AI infrastructure, those demands become part of the investment case.

Follow one gigawatt through the system ↓

01 The resource calculator

IT capacity sets
the resource bill.

Start with IT capacity. The model applies the average electrical load, adds facility overhead, then estimates direct consumptive cooling water. Change an assumption and follow the effect.

Set the operating case

Base expansion assumptions

Share of IT nameplate power drawn on average. This is different from rented space or GPU activity.

Facility electricity divided by IT electricity. Includes cooling, electrical losses and other overhead.

Liters consumed on site per IT kWh. Cooling water only. Indonesian metered fleet WUE is unavailable.

Change resource prices

PLN I-4 high-voltage rate, Q3 2026. Assumes a qualifying industrial contract. The Q4 published schedule and private contract terms have not been verified.

PAM JAYA industrial-large marginal band, 2025. A Jakarta price proxy applied to consumption. Purchased intake may be higher.

Your resource demand Scenario

Electricity energy charges

Rp9.08 trillion

9.11 TWh per year

Energy charges before taxes, minimum bills, reactive penalties and premium services.

Modeled water cost

Rp75.34 billion

3.50 million m³ per year

Direct consumptive cooling water valued at the selected local price.

The electricity build

Annual electricity
IT equipment Facility overhead

Water consumption at this server load

Horizontal axis: consumptive WUE in L/IT kWh. The selected intensity is included. Electrical loading stays fixed.

Average delivered load1,040 MW
Average cooling consumption / day9,600 m³
Share of 2025 national consumption2.00%

Show the calculation and boundaries

IT kWh = IT MW × 1,000 × average load factor × 8,760 hours.
Facility kWh = IT kWh × PUE.
Water m³ = IT kWh × consumptive WUE ÷ 1,000.

Annual steady-state operation is assumed. A monthly value is the annual result divided by twelve. Zero cooling WUE excludes sanitary use and water consumed in electricity generation. The low PUE and zero WUE preset is an exploratory combination that requires an engineering design check.

Same capacity. Different operating demands.

Low, base and high expansion assumptions at your selected capacity and prices. These are conditional cases, not probabilities.

The commentary below uses the fixed 1 GW research base case. The result panels and resource diagram update with your inputs.

02 Follow the resources

From grid supply
to heat rejection.

Power reaches the servers through substations and electrical equipment. Heat leaves through a cooling system. A recirculating coolant loop can still consume water if the final heat rejection uses evaporation.

Resource flows at your selected operating case

IT electricity / year
Facility overhead / year
Cooling water consumed / year

Electricity

Cooling water

Schematic only. Water numbers show consumption. Intake, recirculation and discharge volumes require a site water balance.

Reading the evidence

Today's national total
is still a data gap.

Indonesia's published megawatt figures use different boundaries. A market estimate, a grid connection and a future campus plan each answer a different question.

Reported market estimate322 MW IT

Greater Jakarta, H2 2025

Cushman & Wakefield's dated operational market estimate. It excludes undisclosed capacity elsewhere in Indonesia.

Derived scenario3.84 TWh / year

Demand implied by that capacity

The current-fleet base case uses 80% electrical loading and PUE 1.70. Efficient new-build PUE should not be applied to the entire existing fleet.

UnavailableMetered
national use

Electricity and water

The reviewed evidence does not establish a complete Indonesian data-center consumption total or an AI-specific share.

03 Where the demand lands

A national percentage.
A local connection.

Annual national energy shares do not establish capacity at a particular substation or island water system. Select a region to inspect the disclosed projects and their capacity definitions.

Indonesia / Project atlas

Greater Jakarta & West Java

Selected region
0Filtered records
0With disclosed IT MW
0IT capacity undisclosed

Made with Natural Earth. Markers identify approximate regions, not facility coordinates. Small islands may be omitted at this scale. The list retains projects whose precise location or capacity is undisclosed.

04 The capacity pipeline

Execution changes
the resource demand.

These scenarios draw from dated Jakarta market evidence and separate projects in Surabaya, Batam and Bintan. They cover a documented nationwide subset. The full Indonesian pipeline remains incomplete.

Annual facility electricity

Three paths through the pipeline

2030 comparison
TWh / yearEndpoint values: 2032

Select a year on the chart or focus a case using the legend.

Annual facility electricity at the IT capacity assumed in service. Fixed research cases from assumptions.yaml. The calculator does not change this pipeline view.

01

Power has to reach the site.

Check connection capacity, transformer headroom and delivery timing. Annual national energy does not establish local grid availability.

02

Water needs a local balance.

Check dry-season allocation, permitted intake, discharge and reuse. A low water bill can still sit beside a difficult supply constraint.

03

Cooling choices have to work together.

Efficient server cooling does not by itself determine outdoor water use. PUE and WUE need to be tested against the actual heat-rejection design.

Sources and reproducibility

Open the model.
Check the evidence.

Every calculator result is a scenario. Reported Indonesian observations, derived estimates, foreign technical benchmarks and analyst assumptions stay distinct in the research files.

    Non-Indonesian water evidence is used only as a labeled technical benchmark. Resource costs exclude land, servers, construction, financing and other operating costs. Research cutoff 4 October 2026.