Extensive research has documented that elevated air pollution increases mortality and morbidity, with estimates reaching 8 million deaths per year. Many of the world’s one billion urban poor face both high ambient concentrations and even higher transient peaks. Should government interventions aimed at improving health prioritize reductions in ambient pollution—for example, regulating industrial emissions—or peak pollution? We conduct a field experiment studying the impacts of reducing a notorious source of peak air pollution exposure—biomass cooking—for three years in an urban environment with high ambient pollution. We collect personal, high-frequency particulate matter and carbon monoxide measurements and extensive quantitative and self-reported health measurements. Cooking increases peak PM2.5 exposure by 125 µg/m3 for the control group, but improved stove ownership reduces this by 52 µg/m3—a sizeable 42% reduction in peak cooking emissions. However, ambient pollution of 37.5 µg/m3 largely negates any impact on average air pollution exposure. The reduction in peak cooking emissions generates a 0.24 standard deviation reduction in short-term self-reported respiratory symptoms. However, we can rule out meaningful improvements in blood pressure, blood oxygen, and a wide array of self-reported diagnoses. Ambient air pollution dampens the health benefits from private technology adoption, and chronic health improvements will likely require government intervention addressing negative externalities through environmental regulation. Still, despite the importance of ambient pollution, the $40 stove generates $86 in annual energy savings and reduces CO2 emissions at $4.9 per ton when factoring in additionality rates, suggesting government subsidies would generate large societal benefits.