Nitrogen (N) fertilizer deep placement has been widely adopted to improve nutrient use efficiency and maize yield in the semiarid regions of northwest China. However, previous studies on optimal fertilization depth have yielded inconsistent results across climate conditions, limiting its practical application. This study aims to determine the site-specific optimal N fertilization depth for spring maize by evaluating photosynthetic growth dynamics, yield formation, and N utilization in two contrasting semiarid regions. A two-year (2021–2022) field experiment was conducted in Dingxi (semiarid drought-prone region) and Jingning (typical semiarid region), with five fertilization depths: 0 cm (D0), 5 cm (D5), 15 cm (D15), 25 cm (D25), and 35 cm (D35). A ¹⁵N-labeled urea micro-plot experiment was additionally conducted to trace N fate. The results demonstrated that, compared with the conventional N placement treatment (D15), D25 increased soil total N storage, net photosynthetic rate, root bleeding rate, and the concentrations of NO₃⁻-N and NH₄⁺-N in the bleeding sap. PLS-PM analysis revealed that fertilization depths that are excessively deep (D35) or shallow (D0, D5, D15) adversely affect the photosynthetic parameters and root activity of maize, thereby inhibiting dry matter accumulation and grain N uptake, which ultimately reduces both yield and nitrogen use efficiency (NUE). Compared with D15, D25 increased grain yield, and NUE by 8.79% and 33.19% at Dingxi, and by 7.11% and 11.25% at Jingning. ¹⁵N isotope tracing revealed that D25 improved maize N uptake while reducing residual soil N and N losses. Regression analysis indicates regional differences in the optimal N application depth. To achieve the lowest N residual loss and the highest yield and NUE, Dingxi (23.49 cm) requires a deeper fertilization depth compared to Jingning (21.64 cm). In conclusion, N fertilizer deep placement is a viable strategy for enhancing agricultural productivity and efficiency in semiarid regions, but the appropriate depth should be selected based on local conditions.