Examples described herein relate to improved powering down of PoE interfaces in a network device. In response to detecting a fault that interrupts a flow of power from the PSUs to one or more of the PoE interfaces, the examples determine a total operational power value based on an operational power value of each operational PSU of the PSUs and determine a total available power value based on the total operational power value and a reserved power value. For each of the PoE interfaces, the examples determine whether a threshold power value for the PoE interface exceeds the total available power value and power down the PoE interface in response to determining that the threshold power value for the PoE interface exceeds the total available power value.