Social friction index
How much contestation pressure a data center is likely to face from its surroundings, from 0 (none) to 100.
What it is not: A measure of public opinion (that needs surveys) or of any operator's conduct.
In plain words
- Each data center gets a score from 0 to 100 that describes its surroundings, not the company that runs it.
- Six things are combined: water stress, strain on the electricity grid, how many people live close by, how many other facilities are nearby, restrictive rules in the country, and conflicts already documented within 50 km.
- A higher score means more of the conditions that tend to come before local opposition. It does not predict a conflict.
- The score is relative: 80 means that the surroundings of this data center score higher than those of 80 % of all data centers. The absolute value it comes from, the weighted mean of the six pillars, is shown next to it.
- A country or hub figure is the median of its data centers. With fewer than 10 of them, one site can move it.
Formula
Step 1, absolute value = Σ wᵢ · pillarᵢ / Σ wᵢ over the pillars available for the site. Step 2, index = the share of scored data centers with a lower absolute value, from 0 to 100 (ties share the middle of their ranks). Pressure = the same two steps over the five structural pillars (no conflict data).
How to read it: Since v0.3 the published index is relative. The absolute value (the v0.2 index) is tightly packed: 83 % of data centers fell between 30 and 60 and none above 73, so a score said little without its rank. The index is now that rank: the share of data centers with a lower absolute value. Two sites 10 points apart are separated by a tenth of all data centers, not by a tenth of the scale; the absolute value, kept beside the score, says how far apart their conditions really are. On the map, the five colour classes are fifths of all scored data centers (low, below average, average, above average, high).
Pillars
| Pillar | Weight | Inputs | Normalisation (0–100) | Source |
|---|---|---|---|---|
| Water | 25 % | Baseline water stress, water stress 2050 (BAU), drought risk | mean of the three, each score ÷ 5 × 100 (Aqueduct scores are 0–5) | WRI Aqueduct 4.0 (CC BY 4.0) |
| Energy & grid | 20 % | Data centers per TWh of national electricity demand (60 %), grid carbon intensity (25 %), distance to nearest ≥110 kV substation (15 %) | logarithmic: 0.05 DC/TWh → 0, 15 DC/TWh → 100, clipped; linear: 50 gCO₂/kWh → 0, 800 gCO₂/kWh → 100, clipped; linear: 1 km → 0, 25 km → 100, clipped; weighted mean of available parts | Ember (CC BY 4.0), OpenStreetMap via Overture Maps (ODbL), this inventory |
| Neighbourhood | 20 % | Residents within 1 km (60 %), distance to the nearest residential area (40 %) | logarithmic: 50 residents → 0, 20,000 residents → 100, clipped; closeness = 100 − linear on log₁₀(distance) from 100 m → 100 to 2 km → 0 | JRC GHSL GHS-POP 2020 (CC BY 4.0), OpenStreetMap via Overture Maps |
| Concentration | 15 % | Tracked facilities within 10 km | logarithmic: 1 facility → 0, 150 facilities → 100, clipped | This inventory |
| Policy context | 10 % | Restrictive measures in force in the country (zoning, grid, water and heat-reuse rules, planning restrictions, moratoria) + documented moratoria and grid restrictions. Since v0.2, measures that make siting easier (site designation, fast-track connection, tax relief) and procedural ones (reporting, capacity tenders, non-binding guidance) are not counted, nor are rules whose implementing act is still pending. A country whose regulation has not been researched has no value for this pillar (not 0), and the weights are renormalised | 100 × (1 − e^(−n/2)): 1 measure → 39, 2 → 63, 4 → 86 | Regulation dataset (workbook + researched additions + corrections; snapshot of 2026-09-24, see each country's review date on /regulation/compare, where measures that count are tagged “in friction index”), Conflict atlas |
| Observed conflict | 10 % | Documented contest or mixed events within 50 km (L) and in the country (C); events of the last 3 years weigh 1, older 0.5 | 75 × (1 − e^(−L/1.5)) + 25 × (1 − e^(−C/5)): local events dominate | Conflict atlas (CC BY 4.0) |
Coverage (sites with the pillar): Water 8,330 · Energy & grid 8,348 · Neighbourhood 8,357 · Concentration 8,357 · Policy context 7,055 · Observed conflict 8,354.
How the counts fit together: the inventory holds 8,460 facilities, of which 6,790 are data centers (the others are network points of presence or unclassified). The index is computed at the 8,357 facilities located precisely enough; its maps, rankings and medians use data centers only, 6,687 of the 6,790. The rest have a town-level location or fall outside a source's coverage.
Validation
ROC-AUC of the pressure score (no conflict data) at ranking data centers located within 25 km of a documented contestation (site or municipality precision) above all other data centers. Below: a one-pillar baseline, a drop-one-pillar ablation and a spatially blocked check. They are descriptive: they never change the weights or the scores.
ROC-AUC 0.745 (95 % bootstrap interval 0.728–0.763), 998 positives among 6,687 data centers (0.5 = chance, 1 = perfect).
Baseline: facility clustering only (concentration pillar) scores 0.783. The single pillar does as well as or better than the composite on this test. We publish that rather than tuning weights to win it: the test rewards locations where many facilities cluster near a reported event (press coverage follows clusters), while the index is meant to describe several independent pressures. Read the index as a structured profile, not a better predictor of press-reported conflict.
Drop one pillar at a time
We recompute the pressure score without one pillar (the other weights grow to fill its share) and run the same test. A negative Δ means the AUC falls without the pillar, so the pillar helps on this test. A positive Δ means the score does better on this test without it. The full score has ROC-AUC 0.745.
| Pillar dropped | Weight in pressure | ROC-AUC without it | 95 % interval | Δ vs full | 95 % interval of Δ | Pillar alone |
|---|---|---|---|---|---|---|
| Water | 28 % | 0.736 | 0.716–0.754 | −0.009 | −0.019 to 0.000 | 0.584 |
| Energy & grid | 22 % | 0.733 | 0.715–0.751 | −0.012 | −0.016 to −0.009 | 0.597 |
| Neighbourhood | 22 % | 0.761 | 0.746–0.777 | +0.016 | +0.006 to +0.028 | 0.524 |
| Concentration | 17 % | 0.651 | 0.634–0.669 | −0.094 | −0.101 to −0.087 | 0.783 |
| Policy context | 11 % | 0.715 | 0.697–0.731 | −0.030 | −0.037 to −0.023 | 0.595 |
Dropping Energy & grid, Concentration and Policy context lowers the AUC, so these pillars carry the signal on this test. Dropping Water makes no clear difference (the interval of Δ includes 0). Without Neighbourhood the score does slightly better on this test. We don't re-tune the weights from this table. The test label (a contest reported within 25 km) follows the dense hubs the press covers, so a pillar that describes why people object, such as water or housing nearby, can add little to this test, or even lower the AUC, and still describe a real pressure.
Each pillar of the validated score is dropped in turn; the remaining weights are renormalised (as for a site missing that pillar) and the rounded composite is scored against the same labels. Paired bootstrap, 500 resamples of sites, seed 7: 95 % intervals of the ablated AUC and of Δ = ablated − full. Alone = AUC of that pillar on its own.
Spatial check
Data centers and press-reported conflicts both cluster in a few places. A single worldwide AUC can therefore partly reflect which regions have reported conflicts, rather than how well the score ranks sites. We grouped the 6,687 data centers into 5° × 5° grid cells: 336 cells, of which 27 contain both a site near a reported conflict and a site that isn't. Those cells hold 997 of the 998 positives.
- Within-cell ROC-AUC 0.759 (worldwide: 0.745). Each site near a conflict is compared only with sites in the same cell. The ranking holds up inside cells, so the result is not only a contrast between regions. Only 2 % of the comparisons fall inside a single cell, so this figure rests on the few hubs where most reported conflicts sit.
- Resampling whole cells instead of single sites, the 95 % interval widens to 0.646–0.808 (from 0.728–0.763). It stays above 0.5.
- Other cell sizes: 2.5° cells (620): within-cell 0.710, cell-resampled interval 0.619–0.817; 10° cells (159): within-cell 0.765, cell-resampled interval 0.610–0.790. The conclusion is the same at every cell size tested.
- Cell by cell (21 cells with at least 3 sites of each kind): median AUC 0.672, middle half 0.395–0.887, above 0.5 in 14 of 21. The result varies a lot from one area to another.
| World region | Data centers | Positives | ROC-AUC inside the region | ROC-AUC with the region left out |
|---|---|---|---|---|
| Asia Pacific | 1,178 | 103 | 0.474 | 0.776 |
| Europe | 2,424 | 623 | 0.788 | 0.656 |
| North America | 2,365 | 226 | 0.691 | 0.761 |
| South America | 459 | 46 | 0.919 | 0.732 |
Inside Asia Pacific the score ranks sites no better than chance (AUC below 0.55). Leaving out Europe lowers the worldwide AUC to 0.656, so much of the result comes from there. Regions with fewer than 10 positives or 10 negatives are not shown.
Sites are grouped in 5° × 5° latitude/longitude cells. The within-cell AUC compares a positive only with negatives in the same cell (Σ U / Σ pairs, pooled over cells), so differences between regions cannot count. The cell-by-cell distribution covers cells with ≥ 3 positives and ≥ 3 negatives. The block-bootstrap interval resamples whole cells (500 draws, seed 7). Regions: the atlas's world regions (continents) with ≥ 10 positives and ≥ 10 negatives, scored inside the region and with it held out. The weights are fixed, not fitted, so nothing is trained on the held-out blocks.
Why these weights
Expert judgement, fixed before validation and not fitted to it: water and neighbourhood/energy are the most cited causes of opposition in the conflict atlas (issues: electricity 42, land use 37, water 29 of 80 events), so they carry the most weight; concentration (15 %) captures cumulative impact; policy and observed conflict (10 % each) are context. The policy pillar goes UP when restrictive rules exist, they signal that contestation already reached the political level, not that the country is 'worse'. Changing weights is a new version.
Limits
- Conflicts are documented where press coverage exists. Europe and the US are over-represented; the observed pillar and the validation inherit that bias.
- Clusters of facilities near one event count as several positives in the validation; the spatial check shows how much that widens the uncertainty.
- Energy uses a national proxy (data centers per TWh); official data-center electricity shares will replace it.
- Population is a 1 km grid; facilities with city-level coordinates get approximate neighbourhood values.
Derived indicator of contestation pressure, not a measure of public opinion or of any operator's conduct.
Reproduce & cite
Computed by the atlas's indicator pipeline (version 0.3, run 2026-10-12) from the inputs listed above; every input's source, licence and coverage is in the indicator register. Weights are fixed per version; a change of weights or pillars bumps the version. What changed between versions is listed in the release notes.
Datacenter Atlas (2026). Social friction index, v0.3. https://www.datacenter-atlas.com/social-friction · CC BY 4.0