The issue is as much time as it is money.
$600M is a lot of money and could do a lot of process development. However, it takes time to move material through a factory and then evaluate that material and see what you can learn to make improvements for the next batch of material...Today, Intel has a HUGE advantage here because of the large amounts of money they have spent in the past.
I remember reading some stats on process development costs, and they are going up somewhere between 1.5X and 2X per generation. Another interesting stat is that doing a process node one year earlier approximately doubles the cost. Effectively by doing 45nm a year behind Intel AMD/IBM only need to spend half as much money. If they do it two years it is only a quarter as much money.
This is why scale is currently so important. Intel is up to around $6B/year in R&D spending, so at 100Mu/year that is almost $60/unit in R&D costs. In the last couple of years AMD has grown R&D from about $1B/year to almost $2B/year... and still can't keep up. With around 40Mu/year the $2B/year still comes in at $50/unit... dnagerously close to ASP.
--Alan
Market Data 
Markets 
AI
