Why ? Not enough money in it.
Companies are in business to make money and right now in the graphics chip market that is in embedded devices not pc. In the embedded market there are over 20 graphics chip providers .
I think this is the truth and even reflected in Intel's decision to delay larrabee.
There is a difference, to Intel's decision makers at least and probably any other business leaders contemplating entering the discreet graphics market, between operating at basically break-even (and possibly small losses) with 20-40% gross margins versus being able to confidently project a viable product path for extracting 50-60% gross margins and 10-20% net profits consistently quarter after quarter.
It is simply not enough to look forward 3-4 yrs and conclude "yes we can build a competitive product at a pricepoint that will at best result in us breaking even". Heck it isn't even enough to conclude that you might see 10% net profit on the efforts.
If the projected ROI isn't high enough then the decision makers will cut-bait (projects are internally reviewed on quarterly basis, industry norm) and move to reallocate resources to projects viewed as having potential for higher, and just as important more sustainable, returns on their investment.
The guy operating out of his garage has very different revenue requirements versus the decision makers at a mature business like an Intel or an IBM. He might well be willing to have a business model that calls for operating in the red for a decade at the expense of the venture capitalist's checkbook.
