>SOI is a big gamble, just like the Athlon, Cu and Dresden was before.<
The problem is that AMD lost this gamble, at least with regards to Dresden. The problem was not with copper, but with the capacity it added. Fab30 does not add enough capacity to convince the corporate market that AMD can meet potential future demand yet, provides more than enough capacity to meet retail demand for Duron/Athlon processors (at least as long as no UMA chipsets are available for the production of lowest cost systems), leading to massive price cuts in an attempt to sell out production.
Had AMD not built Fab30, chances are that they would have taken a 16% market share at an ASP closer to $100 than the $75 they actually achieved. Figuring just $90 for the ASP, AMD would have been within $100 million of what their actual revenues were, and not had the expenses of depreciating and operating Fab30.
In hindsight (and not even I would have had the foresight to recommend this), AMD would have been much better off delaying the construction of the Dresden fab by two years. This delay would have enabled AMD to have launched Fab32 in mid-2002 using 300mm wafers. Hector claims that Fab30 alone can produce enough processors to meet 30% of the total worldwide demand for the next couple of years. Even if achieved, this production is not sufficient to provide the corporate market with the confidence they need to commit to AMD, without AMD actually achieving a 30% market share. And, without the corporate market, AMD can not achieve a 30% market share. If Hector is right about Fab30 being able to produce enough processors to meet 30% of the worldwide demand for processors, then this theoretical Fab32, at just slightly higher cost than Fab30, and between 1.6 and 2.2 times the capacity of Fab30, would have sufficient capacity to meet 48% to 66% of total worldwide processor demand. If AMD had sufficient capacity to meet more than half of the total worldwide demand for processors, this would be sufficient to provide the corporate market with the confidence to commit to Hammer. |
