Matt Cutts recently revealed  that Google is now treating subdomains much more like subdirectories of a domain — in the sense that they wish to limit how many results show up for a given keyword search from a single site. In the past, some search marketers attempted to use keyworded subdomains as a method for improving search referral traffic from search engines — deploying out many keyword subdomains for terms for which they hoped to rank well.
Not long ago, I wrote an article on how some local directory sites were using subdomains  in an attempt to achieve good ranking results in search engines. In that article, I concluded that most of these sites were ranking well for other reasons not directly related to the presence of the keyword as a subdomain — I showed some examples of sites which ranked equally well or better in many cases where the keyword was a part of the URI as opposed to the subdomain. So, in Google, subdirectories were already functioning just as well as subdomains for the purposes of keyword rank optimization.
I’ve seen a lot of sites which had varying degrees of quality in their subdomaining strategies. If you do have subdomains, you should ideally insure that they contain primarily unique content not reflected on your other domains — each subdomain should contain page content that does not also live on other subdomains or else it can appear that you are attempting to spam the search engine indices.
Google’s Webmaster Guidelines are very clear on this subject :
“Don’t create multiple pages, subdomains, or domains with substantially duplicate content.”
Most large corporate websites have some level of accidental duplicate content, but if you deploy dozens or hundreds of subdomains with all dupe text, it will appear that you’re purposefully trying to spam the search engines — don’t do it.
If you are considering how to structure your URLs and site content for natural search marketing, I’d say you might be better off just using a simple format of descriptively keyworded directories and subdirectories rather than keyworded subdomains. This is often easier to manage, and it looks a lot more natural/reasonable from the search engines’ perspective. There’s lower likelihood of accidentally mirroring/duplicating your content, too.
Don’t freak out if you have a few subdomains — this is also natural. Many major websites host different site sections and applications on subdomains, and some have external providers delivering content on separate servers — it’s very easy in those cases to assign a subdomain to the third party that’s providing service for you. As long as you’re not duplicating the main content of your pages on the subdomains, this is fine.
Finally, I’ve had a number of people ask my opinion regarding foreign languages — which is better, subdomain or subdirectory.
I actually prefer using separate top-level domains (“TLDs”) for this purpose, since it allows you to send a very clear signal to the search engines that particular content is intended for various countries. For instance, your French language pages could be delivered on .FR domains like: www.example.fr
However, if for some reason you don’t wish to use foreign TLDs for your alternate language pages, you should not worry overly about using separate subdomains versus directory/subdirectories. “french.example.com” will likely function just as well as “www.example.com/french/” in my opinion. I believe that translated versions of pages are NOT counted as duplicate content because they essentially contain very different text. Yes, the information may be duplicated, but the text content is not, and pages in two different languages are far less likely to both come us as relevent for the same keyword search.
So, for foreign language pages, I recommend separate TLDs for best performance, or else use whatever approach is easiest for you to set up and maintain.