You can't access the results individually, and that's the catch. QC does not lead to an exponential speed-up because, even though the "model" says every computation is performed simultaneously, you can only access ONE of the result. As soon as you read one result all the others "collapse" to that result. Imagine a 256 pages book containing the square root of the first 256 integers. Then the only thing you can do is randomly open the book to any page. Say you get sqrt(4)=2. Then every other page of the book holds a 2. You can still use QC to an advantage, but you have to be tricky: waiting as long as possible to read the output and making "interference" operations to increase the probability of the desired answer to show up. That is how a QC can theoretically search an unsorted array in a time proportional to SQRT(n).