Moore's Law predicts that computation will get faster/cheaper by a factor of 2 every 1.5 years; this means a factor of 1000 in 15 years. What will we do with all that cheap computation? Traditional applications of computers (with a few exceptions, such as large simulations) will not absorb enough additional computing power to make them major growth areas.
In the past, computation was scarce and manufacturers could just sell cycles. Now that cycles are cheap, ease of use is a major selling criterion.
Integration of computers with sensors and effectors will be common and essential for competitive position in many markets.
Consumer markets will not be similar to existing computer applications, but will require intelligence to deal with poorly structured tasks (such as cleaning the house).