Abstract Previous studies disagree on whether tropical cyclogenesis is significantly modulated by convectively coupled Kelvin waves (KWs), partially due to limited observations over the satellite era. This study investigates interactions between tropical cyclogenesis and KWs using 10 simulations over 2001 to 2010 produced by the deep‐learning model Ai2 Climate Emulator version 2 (ACE2) trained on reanalysis data. The 100 years of simulations show a robust modulation of tropical cyclogenesis by KWs in all basins. Tropical cyclogenesis preferentially occurs after the peak KW precipitation by 1 day in the Indian Ocean and Western Pacific, mainly due to the enhanced column relative humidity and horizontal vorticity by the KWs. This study demonstrates that large‐ensemble data generated from deep‐learning climate emulators can help strengthen relationships difficult to detect in observations alone. While we focus on tropical cyclogenesis, this approach may also benefit other types of weather or climate variability.