In subsurface drip irrigation (SDI) systems, water constraints to cotton (Gossypium hirsutum L.) production are greatly minimized and N management becomes the main priority. Injecting N fertilizer into SDI systems should, in theory, be as efficient as the irrigation delivery system itself. The objective of this study was to test duration of N injections, addition of the nitrification inhibitor ammonium thiosulfate (ATS), and using canopy spectral reflectance to guide N injections in a 2‐yr study of SDI cotton in Lubbock, TX. For a 1400 kg ha−1 expected lint yield, the N fertilizer rate was 170 kg N ha−1 minus the 0‐ to 0.6‐m soil NO3–N and an estimate of irrigation water NO3 (20 and 30 kg N ha−1 for 2005 and 2006). We tested injection of urea–NH4NO3 (UAN) between first square and early bloom, and first square and peak bloom. There was no effect of timing or ATS on lint yield. Reflectance‐based N management up to peak bloom resulted in savings of 28 and 17 kg UAN‐N ha−1 for 2005 and 2006, respectively, compared with soil‐test‐based N fertigation. Lint yields were similar between reflectance‐based UAN management (1814 kg ha−1) and UAN injected up to peak bloom based on soil test NO3 (1880 kg ha−1). Nitrogen fertilizer recovery at first open boll was not affected by treatment but was high (62–75%). Flux of N2O + N2 was low (800–2100 g N ha−1) in both years and was not different between fertilized and unfertilized plots. Residual 0‐ to 1.2‐m soil profile NO3 after 2 yr was not affected by N fertigation treatments. Nitrogen fertilizer injections in SDI cotton are utilized very efficiently, and spectral‐reflectance‐based approaches have potential to reduce N fertilizer inputs.