This point doesn't seem to make sense. If you have 25 potential attackers, then for there to be an actual 51% attack, more than half of them (13/25) would have to band together to actually reach 51%. Unless the assumption is that the 25 parties buy chips and sit on them indefinitely without contributing to the network hash rate.
I'd guess they mean "better to have one potential attacker because of the centralization that mining with a unique ASIC would tend to cause, than to have 25 potential attackers because you're vulnerable to an attack by 1/25 of worldwide GPU power (which is normally used for other coins but could be briefly repurposed)".