Displacement, velocity, and velocity-squared records of P and SH body waves recorded at teleseismic distances are analyzed to determine the rupture characteristics of the Denali fault, Alaska, earthquake of 3 November 2002 (M_W 7.9, M_e 8.1). Three episodes of rupture can be identified from broadband (∼0.1–5.0 Hz) waveforms. The Denali fault earthquake started as a  M_W 7.3 thrust event. Subsequent right-lateral strike-slip rupture events with centroid depths of 9 km occurred about 22 and 49 sec later. The teleseismic P waves are dominated by energy at intermediate frequencies (0.1-1 Hz) radiated by the thrust event, while the SH waves are dominated by energy at lower frequencies (0.05-0.2 Hz) radiated by the strike-slip events. The strike-slip events exhibit strong directivity in the teleseismic SH waves. Correcting the recorded P-wave acceleration spectra for the effect of the free surface yields an estimate of 2.8 × 10¹⁵ N m for the energy radiated by the thrust event. Correcting the recorded SH-wave acceleration spectra similarly yields an estimate of 3.3 × 10¹⁶ N m for the energy radiated by the two strike-slip events. The average rupture velocity for the strike-slip rupture process is 1.1β–1.2β. The strike-slip events were located 90 and 188 km east of the epicenter. The rupture length over which significant or resolvable energy is radiated is, thus, far shorter than the 340-km fault length over which surface displacements were observed. However, the seismic moment released by these three events, 4 × 10²⁰ N m, was approximately half the seismic moment determined from very low-frequency analyses of the earthquake. The difference in seismic moment can be reasonably attributed to slip on fault segments that did not radiate significant or coherent seismic energy. These results suggest that very large and great strike-slip earthquakes can generate stress pulses that rapidly produce substantial slip with negligible stress drop and little discernible radiated energy on fault segments distant from the initial point of nucleation. The existence of this energy-deficient rupture mode has important implications for the evaluation of the seismic hazard of very large strike-slip earthquakes.