It depends how you count. The 中华字海 Zhonghua Zihai contains 85,568 characters, as Vlad mentions, but the majority of those are either rare variants of
… more common characters, obscure hapax legomena, or other mutant characters that no well-adjusted person has ever seen or used. And the 異體字字典 Yitizi Zidian has 106,230 characters -- though as the name of the dictionary ("Dictionary of Variant-Form Characters") makes clear, these are mostly variants of more common characters. Imagine an English dictionary that collected every single historical spelling of a word -- "cyning," "kyning," "cining," "kyng," "kynge," "king" -- and counted each as a separate word, and you'll have some idea of how this works. Counting simplified 国, traditional 國, and variants 囯, 囶, 囻, 圀, 𡇄, etc., as seven separate characters may make sense from a certain point of view, but it ignores the fact that the two simplified and traditional forms have been used for millennia (as has 囯), whereas the other forms are stylized calligraphic variants that were never official, and would not even be recognizable to most people. And then on top of variants there are "simplified" and "traditional" forms of characters: 華 and 华 look different enough to count as two separate characters, and 鐵 and 铁 look like two different characters, so it seems like the decision to class them as separate characters would be a pretty easy one -- but the majority of simplified and traditional forms are not really as clear-cut. The simplified 车 came from a quickly written cursive form of the traditional 車, and is instantly recognizable as such; ditto 来 and 來. Is 魚 a different character from 鱼? What about 语 and 語? And once you've figured out how to count or not-count all of these, you run into the question of "Chinese" characters as used outside Chinese -- in Japanese, especially, but also in the old Chu Nom writing system used for Vietnamese. If we're going to count 门 and 門 (and 菛 and 𨳇 and 閅) as separate characters, then why not the Japanese form of the character, which is slightly different? And so on and so forth. So basically the short answer is that given the way dictionaries of Chinese characters collect Chinese characters, the "total number of characters" is as potentially infinite as the total number of possible misspellings, alternate spellings, and typographical errors in English. Whatever the actual number of characters is, only a vanishingly small subset of it is necessary to be a functionally literate person. The figure of 5,000 to 6,000 characters that Vlad gives is decidedly on the high side: PRC studies put the number somewhere between 3,000 and 4,000, and even something around 2,000 characters will probably suffice for most newspaper articles. (MORE) 
