Biases in simulating global sea surface temperature stem largely from inaccurate ocean mixing. Near-inertial wave (NIW) kinetic energy (KE) cascades to turbulent mixing upon wave breaking, and is modulated by mean flows via barotropic energy conversion and refraction. Conventional phase-mean KE conversion estimation obscures the instantaneous peak forward KE cascade, causing large uncertainties in long-term mixing effects. Using global drifter data, we explore phase-dependent KE conversion between NIWs and mean flows, revealing a ‘peanut-shaped’ bidirectional KE exchange with two peaks per inertial period. Notably, the peak mean-to-NIW KE conversion drives twice the turbulent dissipation of the NIW-to-mean KE conversion. Over three decades, mean-to-NIW KE conversion intensified by 48%, likely enhancing global turbulent mixing. This new insight provides a novel theoretical framework, showing that phase-dependent KE conversion are intrinsically tied to mixing, which is essential for improving parameterization schemes in climate models.