Hard economics will prompt further semiconductor consolidation

There was a piece of news on Christmas Eve that underlines the growing dilemma that the semiconductor industry is in. Toshiba decided to pass fabrication of its system chips to Samsung, while it will concentrate on future DRAM developments.

Toshiba was making a loss on its system chip business, so the move is, er, logical.

AMD former CEO Jerry Sanders was famous for saying only “real men have fabs”  – of course the X86 company isn’t a real man any more – its fabrication is now created by GlobalFoundries (GloFo) and TSMC.  

Toshiba architects will continue to develop system chips, but Samsung will make them. According to reports in the Japanese press, Toshiba is also contemplating further moves. Its Oita fab will make low end system chips and it is going to sell its Nagasaki fab to Sony. TSMC and GloFo could end up being beneficiaries of Toshiba’s move too.

You have to have very deep pockets these days to contemplate building semiconductor fabs that incorporate the latest technology.

TSMC and GlobalFoundries for that matter continue to invest in the future, simply because they have to, to survive.   Intel’s got very deep pockets, and Samsung has a considerable war chest too.

In the famous words of Future Horizon CEO Malcolm Penn at a conference in Dresden this year.  “If you don’t like innovation, go make Heinz tomato ketchip,” he told delegates. Analysts don’t expect 2011 to be as buoyant for semiconductors as 2010, but it will still see considerable growth.  Semiconductors are just not going to go away.

Of course, Toshiba, Samsung, ST Microelectronics,RenesasIBM, AMD/GloFo and others have a strategic relationship to cooperate on technology for the future, colloquially called “the club” They realised some time ago that the escalating costs of building fabs meant that it was only by pooling resources that they could survive.

Earlier this year, alliance said that it would cooperate to produce 28 nanometre technologies with so called Gate First technology on a “common platform”.  The idea is to synchronise multiple fabs and to match designs to manufacturing specifications.

The mammoth in the room, which in some way prompted the existence of this club is, of course Mighty Intel, which continues to plough its own furrow with ambitious plans to keep Moore’s Law ticking and tocking for just as long as it is physically possible. Introducing advanced technology means that by the middle of the 2010s, a fab could set you back between $5 billion and $15 billion dollars.  And a billion is a lot of anything.

We expect some more consolidation in the industry. While semiconductors are certainly not going to go away unless there’s a global catastrophe of some kind, the face of manufacturing is surely set to change more.