We present a synthesis of long-term measurements of CO₂ exchange in 2 US Intermountain West sagebrush-steppe ecosystems. The locations near Burns, Oregon (1995–2001), and Dubois, Idaho (1996–2001), are part of the AgriFlux Network of the Agricultural Research Service, United States Department of Agriculture. Measurements of net ecosystem CO₂ exchange (F_c) during the growing season were continuously recorded at flux towers using the Bowen ratio-energy balance technique. Data were partitioned into gross primary productivity (P_g) and ecosystem respiration (R_e) using the light-response function method. Wintertime fluxes were measured during 1999/2000 and 2000/2001 and used to model fluxes in other winters. Comparison of daytime respiration derived from light-response analysis with nighttime tower measurements showed close correlation, with daytime respiration being on the average higher than nighttime respiration. Maxima of P_g and R_e at Burns were both 20 g CO₂·m⁻²·d⁻¹ in 1998. Maxima of P_g and R_e at Dubois were 37 and 35 g CO₂·m⁻²·d⁻¹, respectively, in 1997. Mean annual gross primary production at Burns was 1 111 (range 475–1 715) g CO₂·m⁻²·y⁻¹ or about 30% lower than that at Dubois (1 602, range 963–2 162 g CO₂·m⁻²·y⁻¹). Across the years, both ecosystems were net sinks for atmospheric CO₂ with a mean net ecosystem CO₂ exchange of 82 g CO₂·m⁻²·y⁻¹ at Burns and 253 g CO₂·m⁻²·y⁻¹ at Dubois, but on a yearly basis either site could be a C sink or source, mostly depending on precipitation timing and amount. Total annual precipitation is not a good predictor of carbon sequestration across sites. Our results suggest that F_c should be partitioned into P_g and R_e components to allow prediction of seasonal and yearly dynamics of CO₂ fluxes.