When it comes to computers you need to buy JUST below the absolute best available to REALLY make it worth it. The best of the best that is available is crazy expensive at a super premium, because it's 'the best'. One step down is much more affordable, and nearly as good as 'the best'. Build your own computer to be one step below the best and you can expect it to last 5 years as a computer capable of running everything anyone should be able to produce (software wise). After five years it will still be a fine working computer, but not 'on the leading edge'.
Actually, the key is to be constantly changing components out. It may sound like you're spending a lot of money all the time, but it really isn't. I'll give you an example.
Say he puts this together right now, 500$.
In 6 months, he can buy 4 more gigs (2x2GB) - 50$ by then for the same model.
6 months after that, he can get a Phenom II X4 (since he just bought the X2 now).
6months after that, he can buy a Mobo that'll handle >1600 RAM, still use the Phenom X4 and 8gigs of 1600RAM, but have the next socket available and next RAM speed available.
12months after that, he swaps the 8gigs of 1600RAM to 8gigs of (say) 3200Mhz RAM
6-12 months after that, a Phenom II X8
In this scheme, you've always got pretty much the JUST UNDER and its always evolving. The KEY here is you're only spending 50-100$ per 6month or 12month period, so you never incur a large cost! Once you've got a solid case/PSU the rest is constantly progressing
I've used this scheme for 5-7 years now and I just put my old components into older PCs as Servers, Media Centers for my TV, you name it. Heck, since you're replacing components per 6months, you can recover costs and sell on
EBay... Its the same business model as cars; you buy 6months-18months old and sell @4-5years old. you recover costs for new one, you save costs of fixing cars over 4-5years old, you maintain a standard of quality and "newness," other than starting Capital, which is the initial hit, you save a good deal of cost and frustration later on.