What to Sort On
There are a lot of ways a list can be sorted for users. The trick to picking the right one is to analyze the task the user is faced. In your case, the users are looking for potential purchases and want to narrow the search on some criteria. So the question becomes, what criteria do users use to select purchases and eliminate possibilities?
You may have to research your own specific users, but, off hand, I’ve never heard of anyone thinking something like, “I need to buy a bar stool, so I should go with whatever brand has the most (or least or intermediate number of) products.” Thus, sorting by number of products probably provides no benefit to the users, and only makes it harder to find the right choice on some other dimension.
On the other hand, I have heard of customers thinking, in effect, “I need to buy a bar stool, and I know I like Hay’s style.” Thus, sorting brands alphabetically probably provides a benefit to the users when there’s brand awareness, loyalty, and/or reputation.
In general sorting alphabetically is good when, in the task, the user already knows the name of specific item they’re looking for. Then they can quickly scan the list to select the item they want.
Other Dimensions, Other Options
There may be other criteria that are relevant in your case which you could sort brands on, such as length of warranty, quality of service, customer ranking, or popularity, but these probably should be distinct dimensions of the products in a faceted search, rather than confounded with brand.
However, you could, of course, have a little of both in the same list by placing a few selected brands at the top of the list, and then follow it by a divider then a complete alphabetical list. This is a good strategy when a few brands account for most of the filtering actions (check your logs), but you still want to provide a filtering capability on all possibilities.
Showing Number of Products
This is a good practice regardless of the criterion or sort order because it helps the user anticipate what additional filtering they may need to do. Seeing that choosing Vitra will result in 2334 possibilities prepares the user to pick another criterion to make the search manageable. Seeing that, after picking Muuto, the number of products with 2+ years of warranty is 0 tells users not to even bother picking that (and may lead to re-thinking the choice of Muuto before going any further).
Yes, it’s good practice, but not critically necessary. If updating such numbers results in unacceptable server delays, then maybe you’re better off without it. Users will see how many products they get once they click the criterion.