Until last week, the question of whether global warming was speeding up couldn't clear the statistical bar. Natural variability — El Niño cycles, volcanic aerosols, solar fluctuation — muddied the signal enough that acceleration hovered below 95% confidence. A study published March 10 in Geophysical Research Letters by Grant Foster and Stefan Rahmstorf strips those factors out. Across all five major global temperature datasets, acceleration clears greater than 98% certainty: warming over the past decade has reached approximately 0.35°C per decade, nearly double the previous rate of just under 0.2°C.

Zeke Hausfather of Berkeley Earth agrees with the basic conclusion and the approximate magnitude, while noting the methods for removing natural variability are "decidedly imperfect." No one in the field disputes the direction. The argument is over how much. A 2024 study in Communications Earth & Environment argued the data didn't show clear acceleration; Foster and Rahmstorf's methodological innovation — isolating the human signal from natural noise — is precisely what pushed past that threshold.

What caused the acceleration remains genuinely unsettled. The leading hypothesis points to aerosol reduction from cleaner shipping fuel and tighter pollution regulations: fewer reflective particles, more solar energy reaching the surface. But cloud feedbacks and other factors haven't been ruled out. Rahmstorf is explicit: the study identifies that acceleration is occurring, not definitively why. The detection holds across all datasets. The explanation is still being worked out, and those are different scientific questions at different stages of resolution.

What "faster" means where it's already hot

Every major system in Phoenix was calibrated to the old rate.

The city's Heat Response Plan, first passed in 2022 and updated annually, draws its temperature context from NOAA's official 1991–2020 U.S. Climate Normals. Those normals were calculated before the acceleration Foster and Rahmstorf detected, and before the record summers of 2023 and 2024. Every threshold in the plan, every resource trigger, every staffing assumption is anchored to a climate that no longer exists.

The mortality data are specific about what that gap costs. Maricopa County recorded 645 heat-related deaths in 2023. In 2024, the number fell to 608, the first year-over-year decline in a decade, despite 113 consecutive days at or above 100°F. Expanded cooling infrastructure helped. But look at where people died.

Over half of all 2024 heat-related deaths occurred on days the system classified as moderate risk. Ordinary hot days, one after another, the kind the plan treats as manageable. Among those who died indoors, 88% had an air conditioning unit in the room. Seventy percent of those units weren't running. An AC compressor that has operated continuously for weeks doesn't announce its failure with an alarm. It just stops. The room begins warming immediately. At 3am, in a body that hasn't cooled below 97°F — Phoenix's new record for its highest daily low temperature, set in July 2023 — the difference between functioning and non-functioning AC is the difference between a difficult night and a lethal one. The moderate days are where people die.

Grid data track the same pressure through different infrastructure. Arizona's three major utilities set new peak demand records for three consecutive years. In summer 2025, all three exceeded their own pre-season forecasts. APS has filed to raise residential rates by 14%, a direct financial transmission of thermal forcing into household budgets. Ninety-three percent of surveyed Maricopa County schools modified outdoor recess in fall 2023, averaging 3.5 weeks of disruption, under policies designed for a climate where extreme heat was episodic.

Every planning horizon just got shorter

The acceleration compresses every planning horizon. Maricopa County faces a projected 39% increase in days above 110°F by 2050, according to Cotality, a figure derived from models that may not yet incorporate the faster rate. If proposed data center facilities are built, APS and Salt River Project could face up to 17,000 and 12,000 megawatts respectively in new demand by 2038, more than doubling current peak capacity. The grid that already can't predict its own summer is being asked to plan for a load that assumes cooling demand will keep climbing at the old slope.

And 0.35°C per decade is a global mean. Phoenix sits inside an urban heat island that amplifies warming above that floor, particularly at night. The nighttime low is the number that matters most to a sleeping body, because it determines whether core temperature can reset before the next afternoon. That window is narrowing at a rate the plans weren't drawn for.

Heat response programs credited with the 2024 death decline were funded substantially through federal pandemic relief money from the American Rescue Plan Act. That funding expires in 2026. The acceleration study arrives in the same year the money runs out.

The rate changed. Somewhere in Phoenix tonight, a compressor is running.