This paper presents the first application of a digital-intensive intrinsically linear digitally-controlled class-E technique in a Doherty configuration. By careful nonlinear segmentation and multiphase RF-clocking along with overdrive-voltage control and automatic duty-cycle correction, it is shown that even the nonlinearities related to Doherty operation can be fully handled by the underlying design such that digital predistorion DPD can be in principle omitted. The nonlinearity behaviour of the whole digital Doherty PA is analyzed and closed-form equations are given to predict the AM-AM and AM-PM curves. In addition, time/phase mismatch between the Peak and Main branches and the AM and PM signals are accurately compensated. In order to achieve maximum intrinsic linearity, two separate chips with the same architecture, but different design parameters, are fabricated as the Main and Peak amplifiers in 40nm bulk CMOS. To achieve a large RF bandwidth and high passive combiner efficiency, a differential low-loss, wideband Marchand balun-based Doherty power combiner, implemented using re-entrant coupled lines with independent 2 nd harmonic control is proposed, and together with the matching network is fabricated on a two-layer PCB. The measured peak / 6 dB power-back-off P OUT , drain efficiency / power-added efficiency at 2.4 GHz are 17.5 / 12.2 dBm, 57 / 52 % and 36 / 25 % with VDD Main / Peak = 0.6 / 0.7 V. Measured results without using DPD show-41 dBc ACPR and-36 dB EVM for a 16 MHz OFDM signal at 2.5GHz. By using DPD, the measured ACPR and EVM of a 16 / 32 MHz OFDM signals are-52 /-48 dBc and-50 /-48 dB respectively.