I think it still depends on some factors. If the list just groups some records, just to list it out, and the precise order is not important, as users will not go through it in search for particular elements, I think that it does not matter that much. The same if there are just a few elements, so that deep searching is not necessary.
However, once the need for going through an ordered list appears, I would stick to the first model, as it presents the records in a way that is superior regarding comparing records one by one. Let's say you need to go through a list of names which are arranged alphabetically; it will be way more easy for you to find a name while going through each column from top to bottom, as you can focus just on the first letters of the words, which will be more complicated task when sliding with your eyes over the rest of the word, just to skip to another one. Besides, there is a need to provide more spacing between the adjacent records in the second alignment (the horizontal one) which makes it more probable that the user will slip from one line to another, thus making errors.
There is, however, an important disadvantage of the first approach - it can result in very long lists that will need scrolling and then going back to the top of the next column to continue. Thus I think the column height should be limited (in ideal situation: adaptively to the screen) so that scrolling is not necessary while going from one chunk of the list to another. It also helps with finding desired records, as the top row of the columns can be treated as keypoints, or anchors ("is the searched item after this one or before?" - once user gets to a place when it is before, he or she goes to the previous column to refine search).
So, my opinion is:
A E I
B F J
C G K
D H L
--fold--
M Q U
N R V
O S W
P T X
--fold--
Y Z
I think this is the approach learned (and so: expected) by the users from e.g. from dictionaries or atlases.