Does it matter if the cost of running fabs is more in Europe than Asia? It obviously would matter if fab is to be the major value-add of the chip business but, like production operations in many other industries, it’s R&D and design which increasingly make up the value-add.
A hundred square millimetres of silicon could be a $5 memory, or a $500 FPGA with transceivers, special purpose DSPs etc etc on-board.
But printing the hundred square millimetres of silicon costs pretty much the same either way.
So it's not the printing process per se which delivers the value. It's what's printed that constitutes the value. Much like printing a book.
When the Japanese knocked the Americans out of the memory business in 1985, in a way they did the Americans a favour because the US firms then focussed on high-end, high-value, high-functionality chips which carry much better margins than memories.
While the Japanese struggled on for over a decade making losses, or thin profits, in memories.
So, so long as there are enough Asian fabs to keep the printing of silicon wafers competitive, it should not matter too much to Europe and the US if the fabs are all in Asia.
And it looks as if all, or nearly all, the fabs are going to Asia.
As Infineon boss Wolfgang Ziebart pointed out this week, out of 45 fabs under construction in the world, 40 are in Asia. Ziebart attributes that to higher costs in the West.
No one thought tuppence about it when the chip business put its assembly shops in Malaysia in the 1970s. So why should anyone worry about putting fab in Asia in the 2000s?
At the end of the day it’s intellect and imagination that count in the chip industry. Fab was very important when it was a Black Art. But it’s well understood these days.
And where does the major value of a fab come from? From the tools, of course. And who makes tools? Well, Applied and ASML do pretty well at it.
And where do the intellect and imagination come from? From the researchers and the designers, and they're in Munich, Leuven, Sophia Antipolis, Bristol, Cambridge and all points West.
Running a fab is not simply a case of buying the tools from ASML, Applied and pressing a button to make IC's. Considerable intellect, expertise, time and investment are required to integrate the processes that run on those tools in order to get a working product at the end of the line.
Designers working at arm's length from fabs lose the ability to understand the limitations of the processes that they are ultimately working with. Those designers who are closely associated with the fabs are those who will produce the best products. Chip designers in the East can learn from the wafers running and the chips that come out of their fabs.
I believe ASML's prototype EUV lithography tools went to R and D fabs in Leuven and Albany respectively, not to fabs in the East. But without a manufacturing base in the west there will be no-one qualified to run these systems. Where the new fabs go, so the designers and tool manufacturers will eventually follow.