The problem is the ambiguity of the statement. Is it 50% of each species? Or is it 50% of all life as one set?

If it’s the former case, then sure 3.5b humans and n/2 bacteria gets snapped.

But if it’s the latter case, we group all 7b and n bacteria into one set and snap half of them. This 50% can consist of 50% humans + 50% bacteria, but there’s also a chance for it to include 0% humans + 100% bacteria. Therefore, the amount of humans snap is a random variable instead of a constant.