I see that John Dvorak is blaming Intel's Itanium for having "Killed the Computer Industry". That seems like a premature obituary: overall PC unit sales in 2008 were up more than 10% versus 2007, with Gartner and iSuppli both projecting further growth (albeit only around 4%) in 2009.
I can think of plenty of people who'd be thrilled to be having that kind of "collapse" (another of JCD's hyperboles).
More interesting, to me, is the question of cause and effect — and which is which. Dvorak asserts that the Itanium's advance hype killed off all the other promising directions in processor innovation, and that programmers are still stuck with the ancient x86 instruction set as a result.
We can't very well go back and run the experiment again, but it seems more likely to me that hardware people thought that they were still in charge of defining the industry's future — at a time when the inertia of the worldwide base of programming skills was actually the overwhelming factor. If programmers don't know how to write correctly sweetened code, a processor's theoretical strengths are irrelevant.
Learning how to craft outstanding code for a new machine, with or without the help of a high-level language, is a skill that costs real money — time spent, tools bought, training endured, opportunity costs incurred — and developers need a compelling reason to make that expenditure. A mere factor of two, say, in performance is barely enough: by the time the code quality is what it needs to be, hardware price/performance improvement would likely have gotten you to the same place.
At least, that's the way it used to be — but there's a real question of whether hardware improvement is still enough, unless programmers also embrace the techonomics of the cloud.
Managed code on managed platforms represents another big opportunity/challenge for developers, but I believe that this time — unlike the sinking of the Itanic — the developers will be willing to dive in and learn to swim. The cloud is where it's faster to go to market with a better product that more customers can discover, evaluate and use in less time at less cost.
Multiply all those factors of 2 (time to market), 1.5 (product usability), 1.8 (adoption) and 1.33 (favorable cost ratio) and you get something close to an overall factor of 7. Pick any reasonable values you like, you'll still come up with two or three Binary Orders of Magnitude (factors of 2, to those outside our little world). That's a 3- to 5-year jump ahead of any competition that's still writing code for single-user chunks of bare silicon.
That's a competitive edge that's finally worth what it costs to achieve it.