Nowadays, most computers use Little Endian, however, that might change in the future.
Also, there's no justification to write software only for one endianness due to speed(tm) or something. Learn to use the interfaces provided by the system (htons(),htonl(),ntohs(),ntohl(),...)
and when you specify interchange formats, specify an endianness.
Which one it actually is doesn't matter, there's not too much magic happening in the background anyway. Those people claiming "Little Endian Master Race" might probably jump on the Big Endian bandwaggon as soon as it became the majority for some reason.
In fact though, writing endian independent code is the way to go.