The remarkable sensitivity, frequency selectivity, and nonlinearity of the cochlea have been attributed to the putative 'cochlear amplifier', which consumes metabolic energy to amplify the cochlear mechanical response to sounds. Recent studies have demonstrated that outer hair cells actively generate force using somatic electromotility and active hair-bundle motion. However, the expected power gain of the cochlear amplifier has not been demonstrated experimentally, and the measured location of cochlear nonlinearity is inconsistent with the predicted location of the cochlear amplifier. We instead propose a 'cochlear transformer' mechanism to interpret cochlear performance.