An equalization training method and apparatus are described. The method includes obtaining a training rate of each of a master chip and a slave chip in a target phase of equalization training. The method also includes determining a target rate threshold interval within which the training rate in the target phase falls, determining, based on a correspondence between N+1 rate threshold intervals and N+1 equalization timeout periods, a target equalization timeout period corresponding to the target rate threshold interval, and configuring the target equalization timeout period as an equalization timeout period in the target phase. According to this method, an equalization timeout period used for equalization training can be flexibly configured for each equalization training process, so that the configured equalization timeout period better conforms to a training rate currently used for link negotiation, to better ensure that an equalization parameter is found within the configured equalization timeout period, thereby improving an equalization training success rate.