Satellite observations and energy balance constrain the thermal footprint of a hyperscale data-centre cluster

Jiawei Jing 

Whether hyperscale data centres (DCs) warm their surroundings divides current evidence: a coarse-resolution satellite preprint attributes ~2 °C of land-surface-temperature (LST) warming to facilities worldwide (Marinoni et al 2026), while peer-reviewed transects near Phoenix measure +0.7-0.9 °C of near-ground air warming a few hundred metres downwind (Sailor et al 2026). Prospective anthropogenic-heat models remain untested against long-term observations, and no major DC market assesses local temperature impact before construction. We provide the first ex-post, controlled satellite test for Japan's largest DC concentration, the Inzai and Tama clusters near Tokyo, whose staggered build-out (twelve facilities, 2011-2025) yields a three-phase event study with replication. The test combines daytime Landsat (2013-2026), 70 m day/night ECOSTRESS and 1 km MODIS with matched pre-2013 controls, facility-level energy disclosures and minimum-detectable-effect analysis, gated by an air-temperature energy-balance ceiling (~0.35 °C, cluster scale, night) that bounds LST a fortiori. Commissioning cools the daytime LST by 1.8-2.2 °C — significant under all three reference types, a radiative-property signal opposite in sign to the viral claim. Deep-night facility-scale LST changes are indistinguishable from zero (bounded below ~0.6-1.2 °C), and the cluster-scale night LST bound (0.24-0.29 °C, autocorrelation-corrected, two platforms) closes beneath the air-temperature ceiling. The exclusion of warming at the claimed magnitude rests primarily on energy balance — +2 °C, even in air, would require 6-43 times the available waste heat — with observations supplying consistent bounds at every scale they resolve. Arid- and continental-climate consistency checks (Phoenix and DeKalb) reproduce the decomposition: in both, night surface warming is largely in place during construction — before any server load — and matches neighbouring non-DC development; conversion, not exhaust. Near-source air warming is real where rejection density is high and climates dry; kilometre-scale surface warming from waste heat is unattainable. The protocol is the ex-post verification tool siting regimes lack.