Background To evaluate ultra‐short‐echo‐time (UTE) MRI pulmonary signal‐intensity measurements and reproducibility in chronic obstructive pulmonary disease (COPD). Methods A two‐dimensional sequence (echo‐time = 0.05 ms; acquisition‐time = 13 s) with interleaved half‐pulse excitation and radial ramp‐sampling was used with compressed‐sensing to reconstruct UTE images from under‐sampled data. Five healthy volunteers and 15 subjects with COPD provided written informed consent to imaging and pulmonary‐function‐tests. Healthy volunteers underwent MRI at four lung volumes: full‐expiration, functional‐residual‐capacity (FRC), FRC+1L, and full‐inhalation; COPD patients underwent computed‐tomography (CT) and MRI at FRC+1L. Three‐week reproducibility was evaluated and the relative area of the density histogram ≤ −950 HU (RA950) was compared with mean MRI signal‐intensity. The 15th percentile of signal‐intensity‐histogram (SI15) was compared with the 15th percentile of the CT‐density‐histogram (HU15). Results In healthy subjects, signal‐intensity correlated with the inverse of lung volume (r = 0.99; P = 0.007). Contrast‐to‐noise and signal‐to‐noise ratios were significantly improved for 32‐channel UTE (P < 0.01). The coefficient of variation for 3‐week repeated measurements was 4%. There were significant correlations for signal‐intensity with RA950 (r = −0.71; P = 0.005), FEV1/FVC (r = 0.59; P = 0.02), and for SI15 with HU15 (r = 0.62; P = 0.01). Conclusion Pulmonary signal‐intensity is reproducible and related to tissue density. In COPD subjects with and without bronchiectasis, signal‐intensity was also related to pulmonary function and CT measurements. J. Magn. Reson. Imaging 2015;41:1465–1474. © 2014 Wiley Periodicals, Inc.