Cikarang, east of Jakarta, holds most of Indonesia's new data center capacity, on the same alluvial plain that has made North Jakarta one of the fastest sinking cities on record. We measured the ground under eight data centers from the free Sentinel-1 radar archive: 61 satellite passes between September 2024 and September 2026, 119 interferograms, sampled at building coordinates taken from PeeringDB and OpenStreetMap.
Every one of them is flat to within our own measurement noise. Find your site in the table below. The full working, including a mistake we made and corrected in public, is further down the page.
Rate of line of sight movement over the two years, from the median of the radar pixels within 400 m of each coordinate. The ± is one standard error, and a site reads as flat when its rate sits inside about two of them. Click a row to pick it out on the chart.
| Data center | Estate | Rate, mm/yr | Pixels | Coordinate from |
|---|
One point per industrial estate, sampled in our first pass. They are points on the ground, not buildings, and not a verdict on a whole estate.
| Point | Note | Rate, mm/yr | Pixels |
|---|
Median line of sight displacement inside 400 m of each data center, one point per satellite pass, relative to a reference strip south of the estates. Down means moving away from the satellite. Click a name to hide it.
Line of sight velocity across the scene, red moving away from the satellite. The small circles are the verified data centers: DCI H1 in MM2100 to the west, EdgeConneX JKT01 in the middle, and six more in the GIIC cluster at Kota Deltamas to the south. The orange circle is the estate point we first sampled, on the southern flank of the subsiding area that covers the north east of the scene. That area looks like regional groundwater drawdown, not load from anything built on top of it.
Nothing here clears its own error bar, and that is the finding. Free radar over two years puts the ground under all eight halls inside a few mm/yr of each other and of zero. An operator who was about to commission a ground survey on a hunch now has an ordering: none of these eight, and a subsiding area 6 km north east that would matter a great deal to anyone building in it.
What it is not is a certificate. Free radar over two years can tell flat ground from moving ground, but on tropical ground it cannot give an exact rate. When we checked our own rates against GNSS stations on tropical ground in Puerto Rico over six months, they were off by 54 mm/yr RMS, which spread across two years is still of order 20 mm/yr on an absolute rate. Nothing in the table clears that bar either, which is why this reads as a null and not as eight precise measurements.
The first version of this study sampled one point per industrial estate. The point for Kota Jababeka, an estate of about 5,600 hectares, read 23 mm/yr faster sinking than its neighbours, and we named the estate's data center tenant beside the number. That point sits 6.2 km from the nearest verified data hall, on the flank of a regional subsidence area. The ground at that point is moving, and no data center is on it. How much of that area falls inside the estate boundary is something we cannot see from outside.
The middle button on the chart puts that series back. It is the same pipeline, the same stack and the same inversion, sampled 6 km away, and it is a 23 mm/yr difference in the headline. That is why every study now starts from the coordinate of the building.
Search for your site or click it on the map. This checks the real satellite record over that exact spot: the passes, the orbit, the freshness. Give it the coordinate of the building, not the name of the park.
The same pipeline runs over a data center campus, a tailings dam, a slope, a reclamation site or a port. Name two sites and we will run the free archive over them at no cost, and send you the same record as this page.
Disclaimer. Rheality provides decision-support monitoring data, not a safety guarantee. The duty of care for any asset remains with its owner and their engineer. See Terms & Disclaimer.