Big-O is a way to talk about scaling, but in general, the average CS program does a disservice by not talking more about the critical question: "of what?" Bright students will connect their algorithm course with their systems course and recognize that, maybe, ALU ops aren't a very useful measure in a real world context. Interested students may take a rigorous complexity theory class that will get into various computation models and their nuanced interaction with time/space/communication complexity. However, enough students seem to go through a CS program without learning this that I'm convinced this is a failing of the curriculum and not the students.