You know how speed cameras work on highways? They don't stop you from speeding — they catch you after the fact, with a timestamp and a number that's hard to argue with. This paper turns a geostationary satellite into a speed camera for air pollution, and the first facility it clocks is doing 16× the limit. The committed claim: space-based NOx detection and quantification from a single AI data center power plant has never been done before. Gauld et al. use NASA's TEMPO instrument — a geostationary spectrometer measuring NO₂ column concentrations over North America — to detect and track emissions from the SpaceXAI Colossus 2 natural gas turbine plant in Southaven, Mississippi. TEMPO first picked up the signal in December 2025 at 460±180 kg h⁻¹. By February 2026, emissions averaged 730±185 kg h⁻¹ — roughly 16 times higher than the ~47 kg h⁻¹ expected from the facility's March 2026 permit covering 41 turbines under best available control technology (BACT) requirements. The ladder here matters. Ground-based continuous emission monitoring systems (CEMS) are the current gold standard for power plant NOx tracking, and the authors calibrate against CEMS data from other US power plants. They're not claiming TEMPO beats CEMS on precision — they're claiming it can see things CEMS can't, because CEMS requires the emitter to install and operate the monitors. The whole point is that TEMPO provides independent, non-cooperative verification. At the expected BACT-compliant emission rate of ~47 kg h⁻¹, the signal would be undetectable by TEMPO — meaning the satellite only caught this because the facility was massively exceeding its permit. Architecturally, this is remote sensing retrieval applied to tropospheric NO₂ columns from a UV-Vis geostationary spectrometer. The key structural advantage is TEMPO's geostationary orbit — unlike polar-orbiting satellites like TROPOMI that get one overpass per day, TEMPO provides multiple daily scans of the same location, enabling two-week averaging that pulls signal from noise. The calibration against CEMS is the methodological bridge that converts column concentrations into source-rate estimates (kg h⁻¹). Integrity is solid for what this is. The authors use an independent, operational satellite instrument (TEMPO, launched 2023) and calibrate against EPA-mandated CEMS data from other facilities — not the facility in question. The before/after design (August 2025 pre-operations baseline vs. post-operations) provides a natural experiment. The 16× exceedance is so large it swamps the uncertainty bars (730±185 vs. ~47 expected). Cherry-picking risk is low: they're measuring one facility over a continuous time series, not scanning the country for outliers. The milestone trajectory is clear. TEMPO can currently detect sources above roughly 400-500 kg h⁻¹ NOx. The next threshold is detecting facilities at their permitted levels — which for BACT-compliant gas turbines means pushing detection sensitivity down to ~50 kg h⁻¹, roughly a 10× improvement. That probably requires either next-generation geostationary spectrometers or multi-satellite fusion. Until then, TEMPO is a tool for catching gross violators, not routine monitoring. The obvious successor experiment the authors didn't run: applying this method systematically to every known AI data center gas turbine installation in the US. There are dozens coming online. My read is (c) — they're saving it. This paper establishes the method on a single, dramatically non-compliant facility. The survey paper is the follow-up, and it will be more impactful if regulators are already paying attention to this one.
