The gamma-ray burst (GRB) of 1994 October 17 (GRB 941017) showed a distinct high-energy spectral component extending from ≲a few to ≳200 MeV, in addition to the typical GRB emission, which peaked at ≲a few hundred keV. The high-energy component carried at least ∼3 times more energy than the lower energy component. It displayed an almost constant flux with a rather hard spectrum (Fν ∝ ν-α with α ∼ 0) from ≲20 s into the burst up to ∼200 s, while the duration of the GRB, where 90% of the energy in the lower energy component was emitted, was only 77 s. Such a high-energy component was seen in only one out of ∼30 GRBs in which a similar component could have been detected and thus appears to be quite rare. We examine possible explanations for this high-energy spectral component and find that most models fail. The only emission region that provides the right temporal behavior is the reverse shock that goes into the GRB ejecta as it is decelerated by the ambient medium, or possibly the very early forward shock while the reverse shock is still going on. The best candidate for the emission mechanism is synchrotron self-Compton emission from the reverse shock. Even in this model the most natural spectral slope is only marginally consistent with the observed value, and some degree of fine-tuning is required in order to improve the agreement. This might suggest that an additional or alternative emission mechanism is at work here. A prediction of this interpretation is that such a high-energy component should be accompanied by a bright optical transient, similar to the one observed in GRB 990123.