Could GloFo Cut And Run At Dresden?

Has the penny finally dropped in Abu Dhabi that their move into the silicon foundry industry has been poorly executed?

This week, the backers of GloFo, the investment trust ATIC, have said they’ll not be breaking ground on a new fab in Abu Dhabi next year.

The idea was to have an Abu Dhabi fab out-putting wafers in 2015.

Shivers must have run down a few spines in the Gulf last week when AMD said they’d not be using the GloFo 28nm process.

Reports in the summer said the Dresden 28nm process had zero yield. Quite recently this had risen, so it was said, to one die per wafer.

It’s not the fab or the equipment at Dresden which are the problem – both are first-rate. The problem is the way it’s being run and the choice of the gate first process approach.

Now if TSMC was running that fab do you think it would be getting that kind of yield?

And if TSMC ran Dresden, would AMD decide to go back and have its chips made there?

Well it might. And that would give Dresden a value on which ATIC could capitalise.

Which is why it is being said in some quarters that ATIC may sell Dresden to TSMC and invest the proceeds in the GloFo fab being built in New York.

And in attracting some good people to run it.


Comments

13 comments

  1. I knew some of the managers at Chartered and they always impressed me when dealing with them, though as you say many of them left soon after the takeover.
    I also know some of the AMD design team and they are very good so I really do think the problem lies elsewhere. Whether it is with the process itself or bad communication between AMD and GF isn’t known as GF don’t give any details even when they receive bad press. TSMC did admit they had problems a couple of nodes back and had a plan to fix matters – which they did. Strangely this is the opposite of what one’s preconceptions might be that those in the Far East would be secretive whilst Americans and Europeans would be more open. I suspect this is entirely down to Morris Chang – simply the best CEO in semis at the moment.
    Regarding Qimonda Dresden, this was a memory fab at the depth of a bad time for DRAM so the fab had little value without a major refit – which Infineon are now doing. Conversely the GF Dresden fab is a logic fab at or near the state of the art and would be an attractive buy at the right price.
    Either way something needs doing as the bad press can’t be attracting new customers to GF.

  2. poor fab management in dresden is related to bad managers from chartered after its takeover by GF?
    note that the cream had left chartered during its struggle.
    and how robust is AMD’s design in general so that TSMC will fab out high yields starting from 1st silicon?
    Since samsung or no one for that matter acquired the struggling qimonda i dont see any reason why the dresden GF site specifically would be interesting for tsmc/samsung.

  3. Ari, Your comment about “deep pockets” reminds me when I was saying the same thing about UTC after it bought MOSTEK. Then when M started losing a million dollars a day, UTC decided their pockets weren’t that deep! 🙁
    anon

  4. Yes, Stooriefit, Ogilvy was a first rate man, and first rate men hire first rate men.

  5. Interesting documentary on Radio Scotland earlier in the week about David Ogilvy of “Confessions of an Advertising Man” fame.
    He apparently wrote a memo to his managers urging them to hire subordinates who were more capable than they were, even if it meant paying them better than themselves.
    He certainly seemed like an interesting fellow and built Ogilvy & Mather from virtually nothing so must have had a something special.

  6. That’s certainly my understanding, Anonymous. As they say, second rate men hire third rate men, third rate men hire fourth rate men and fourth rate men hire morons.

  7. Maybe the answer lies in that comment ‘world+dog’; dogs are certainly what they’ve hired … not a semiconductor visionary amongst them 🙂

  8. Well, Anonymous, there was the removal of the CEO back in June (allegedly over poor yields at Dresden) followed by reports of zero 28nm yields at Dresden in the summer, followed by reports of one die per wafer yields recently, followed by AMD announcing that it wouldn’t use GloFo’s 28nm process, followed by ATIC’s announcement that it wouldn’t start building a new, planned fab in Abu Dhabi. Does that look like a pattern to you? It suggests growing disillusionment to me.

  9. Well it would be bad news if ATIC turn shy with regard to funding. Building fabs and hiring world+dog is pretty costly … probably not covered by current revenue stream?

  10. Oh absolutely not up for sale officially, Daniel, but don’t you detect an air of disillusionment at ATIC HQ? And your remark about Samsung’s 28nm might explain that feeling.

  11. I think Samsung would be a more likely buyer for Dresden. Samsung is yielding with the same 28nm process recipe that GFI is using.
    I highly doubt Dresden is for sale however. In fact, GFI is said to be looking at buying a fab in Taiwan. That would make much more sense than building a fab in the Middle East.
    D.A.N.

  12. Because, Aristotle, ATIC is reducing its investment in GloFo – e.g.shelving/cancelling the Abu Dhabi fab – and, if they want an exit or partial exit strategy from GloFo, there’s no one else who’ll buy Dresden.

  13. Why would GloFo sell anything to its primary competition? … ok it might get some cash, but ATIC has deep pockets – surely it MUST get the technology in Dresden working to win.
    Why would TSMC buy something that is perceived not to be working? … maybe, if it was cheap … maybe an agenda for geographical diversification (putting the Global into TSMC in more ways than one)

Leave a Reply

Your email address will not be published. Required fields are marked *

*