> […] characters that aren't in ASCII, like say Chinese, UTF-8 will consume much more storage than necessary.

I don't think that's a problem in practice. Taking for example:

http://www.baidu.com/s?wd=%D0%C2+%CE%C5

The page takes 37KB in GBK, 39KB in UTF-8 and 73KB in UTF-16! (UTF-16 doubles cost of HTML markup but only saves ⅓ in text)

Even in pure Chinese plain text UTF-16 doesn't win by much. I've tested some random article: 18KB in UTF-16, 25KB in UTF-8. It's down to 9.5KB vs 10KB after gzipping.