This paper proposes a stochastic geometry framework to evaluate the coverage and rate performance in millimeter wave (mmWave) cellular networks. The authors model the locations of line-of-sight (LOS) and non-LOS base stations as two independent non-homogeneous Poisson point processes, applying different path loss laws to each. Expressions for the signal-to-interference-plus-noise ratio (SINR) and rate coverage probability are derived, considering antenna geometry and base station density. The analysis shows that dense mmWave networks can achieve comparable coverage and significantly higher data rates compared to conventional ultra-high frequency (UHF) cellular systems, despite blockages. A simplified system model is introduced to analyze dense mmWave networks, where the LOS region is approximated as a fixed LOS ball. The results indicate that the optimal cell size scales with the average size of the LOS region, and increasing base station density does not always improve SINR, suggesting a finite optimal density. The paper also provides numerical simulations to validate the analytical results and discuss their implications for system design.This paper proposes a stochastic geometry framework to evaluate the coverage and rate performance in millimeter wave (mmWave) cellular networks. The authors model the locations of line-of-sight (LOS) and non-LOS base stations as two independent non-homogeneous Poisson point processes, applying different path loss laws to each. Expressions for the signal-to-interference-plus-noise ratio (SINR) and rate coverage probability are derived, considering antenna geometry and base station density. The analysis shows that dense mmWave networks can achieve comparable coverage and significantly higher data rates compared to conventional ultra-high frequency (UHF) cellular systems, despite blockages. A simplified system model is introduced to analyze dense mmWave networks, where the LOS region is approximated as a fixed LOS ball. The results indicate that the optimal cell size scales with the average size of the LOS region, and increasing base station density does not always improve SINR, suggesting a finite optimal density. The paper also provides numerical simulations to validate the analytical results and discuss their implications for system design.