Electronics Weekly Electronics Design & Components Tech News
If it wants to compete with TSMC on cost – and Glofo says it wants 30% of the world foundry market – then it’ll have to manufacture on the same scale as TSMC.
And TSMC has two Gigafabs and has announced the start of a third.
So why doesn’t Glofo build a Gigafab?
Well one reason may be that it wants to get grants from the governments of the countries where it operates.
It is significant that Glofo’s announcement of expansions at Dresden and New York are dependent on getting investment approvals from the GermanState, the EU, the Empire State Development Corp of New York and from the State of New York.
If Glofo want the grants they have to stick to local fab sites, and if they stick to local fab sites they dilute their investment across a stable of sub-Gigafabs and guarantee a higher wafer cost than a TSMC Gigafab.
Recently I asked a Glofo exec – Mojy Chian, Glofo’s senior vice president for device enablement –Â how Glofo could compete against TSMC on cost when all Glofo’s fabs are considerably smaller than a TSMC Gigafab.
Chian evaded the issue, responding instead with a swipe at TSMC’s recent yield difficulties on 40nm.
“In the advanced technology side, cost is not the only factor”, said Chian, “you are all aware of the issues at 40nm. The early adopters at 40nm had issues which were independent of cost. They couldn’t get wafers, no matter what the cost.”
Well, Yes, it’s a point, but it doesn’t even begin to address the real problem.
The saving isn’t really in the actual wafer processing as a giga-fab is just more lines. But you do save on support infrastructure – air cleaning, gas supplies, etc. Also you need less maintainence and other indirect experts as they can be quickly called to any problem – not possible if they are a continent away.
Martijn, I did once see TSMC’s cost/benefit analysis for wafers produced in 100k wpm fabs compared to 30k fabs, but have forgotten wht it was. I’ll ask TSMC for it again. Meanwhile I think it’s accepted that TSMC has the lowest wafer cost, highest profit and best margins in the foundry industry. Maybe that’s related to large fabs. I’ll try and find out
What is the cost benefit of having (much) bigger fabs? I tend to agree with Global Foundries that especially for non-commidity chips, scale of operation is only a limited contributor. Above 30k wspm, going bigger will not save you so much.
Futhermore geographic spread limits other risks for continious output, as the quake in Taiwan last March showed, and gives access to a larger workforce.
Cynical, Mike, but probably accurate
1. Build your fabs in large western economies
2. ‘Discover’ you are being undercut by Far East manufacturers
3. Lobby governments (who are your investors of course) to put up trade tariffs
Simples !!
Well it worked in the late 80s/early 90s 🙂