Could 450mm Cost $40 Billion?

Could 450mm really cost $40 billion? That is what CEA-LETI senior advisor Michel Brillouet, is saying. SEMI’s estimate is $25-40 billion.

Much of the investment will be centred at the Global 450 Consortium (G450C).

Here’s the problem for European equipment and materials suppliers – unlike current 300mm process development which occurs at multiple locations around the globe – near-term 450 development seems likely to be exclusively conducted at the G450C site in New York.

Tom Jefferson, 450 programme manager at Sematech says that, by mid-2013 to early 2014, a complete 450mm production line will be established at Albany containing 50 different tool types.

Here’s where things get tricky. The objective of the pilot line, with its 50 tool-types some from multiple suppliers, will be to develop data to support the purchase of production-line tools. It is unlikely, says Jefferson, that non-participants in G450C will be favourably considered for 450 production lines.

Participants will benefit from access to patterned and non-patterned wafers, shared metrology and Multi Application Carriers (MACs), shared consortium staff resources, data sharing, and “financially leveraged business partnerships” with consortium partners.

Suppliers who do not participate in the programme will be lower on the priority for access to test wafers. The impression left was that not participating in the program will lower the probability of participation in production line rollouts by consortium partners.

“Some equipment suppliers may choose to stay off 450mm and focus on specialty development on 300mm platforms, but the 8nm node is likely to be the 450mm equivalent of 65nm’s ’300mm only’ moment,” says SEMI, adding “furthermore, the ‘Post CMOS’ era will likely be exclusively on 450mm wafers.”

Once 450mm is developed, spare capacity in 300mm will emerge, encouraging a migration from 200mm production, impacting the viability and competitiveness of both 200mm and 300mm fabs, says SEMI. In 15-20 years, even low volume, mature technologies in MEMS, power and analogue could migrate to 450mm fabs.

CEA-LETI’s Brillouet says the EC’s options are: help develop 450 and forget all More-than-Moore projects; forget 450 and concentrate on other process technologies such as 3D, EUV, etc, and support R&D by equipment and material suppliers without a European production or pilot development lab.

How much of the headline $4.4 billion investment in 450mm at Albany represents already-committed IBM money for other (non-450) advanced chip design and technology development is not known.

Tags: consortium partners, consortium staff, globe, line tools, wafers

Related Tech News

12 Comments

  1. Ian Dedic
    November 18, 2011 09:48

    Mike, your comments about Intel/Samsung/TSMC are spot on. But it’s also linked to the fact that these are the big three who can afford mega-fabs and 450mm.
    Bit of a problem for everyone else though, not just the other foundries but also foundry customers who will effectively end up with TSMC as the only supplier — maybe with GloFo as a much smaller second source, like AMD is to Intel.
    Intel’s fabs are optimised for CPUs not SoC, and they make much more money fabbing their own products for sale at high margins than flogging wafers at much lower margins.
    Samsung also make more by selling end products, especially to their own product divisions — they really *are* an IDM in the true sense of the word, where the device is a TV or phone using their own chips made in their own fabs.
    It could be argued that an almost-monopoly situation is inevitable given the continually escalating costs of fabs and equipment, but just think how everyone complains now about Intel’s high CPU prices but carries on buying because there’s no real choice — the risk is that CMOS SoC could end up in the same position.
    Don’t see what can be done though, apart from an international government-backed consortium whose remit is to supply the most advanced technology to enable global business without holding people over a barrel — and I don’t see *any* chance of this happening…

  2. David Manners
    November 18, 2011 08:47

    You’re very much in tune with what the European semi companies want, Greg, they’re the best in the world at power technologies and SiC and GaN are what they want. But that doesn’t keep the European semi manufacturing equipment companies in the game. If they want to be able to supply to 450mm fabs, they’ll need government help.

  3. greg
    November 18, 2011 00:38

    David, those are both better options I think. There’s enough space around GaN and SiC for the UK to stand up something like a Cree with modest investment if you could get the right technology out of the project. Thats a better outcome I think than will be a 450mm project.

  4. David Manners
    November 17, 2011 22:22

    Very true, Mike

  5. David Manners
    November 17, 2011 22:12

    Well the alternative, Greg, is developing processes like Gallium Nitride and Silicon Carbide which the EC is also supporting. And this seems to be what ST wants. The big one, of course, is graphene which the UK government, believe it or not, is spending 50 mill on.

  6. greg
    November 17, 2011 20:29

    This is all partly why the EU 450mm project sucks. Just a public money flush. I’m surprised we can’t come up with something different that pushes competitive technology into up and coming SME’s.

  7. Mike Bryant
    November 17, 2011 19:30

    But of course Intel, Samsung and TSMC aren’t run by accountants. And who are the top players in the industry – Intel, Samsung and TSMC. Some correlation maybe ?

  8. Ian Dedic
    November 17, 2011 17:57

    I doubt somehow that even governments are going to cough up a big enough pile of dosh to cover 450mm development, given that they’re mostly in financial meltdown right now and have higher priority things to spend tens of billions on than cheaper chips a few years down the line, probably made in another country…
    They might provide enough seed money to kickstart it, but the majority of it will still have to come from industry so the problem is still there.

  9. David Manners
    November 17, 2011 16:43

    As you say Ian, it’s an extraordinary thing that, with most semi companies run by accountants, they can even contemplate spending up to $40 billion on a technology which will result in $10 billion factories needing to be built. Which is why even Intel, Samsung and TSMC have waited for a sucker ( I mean government) to come in with a funding offer.

  10. Ian Dedic
    November 17, 2011 15:21

    If you use the normal 4-year period for writing off development costs, it will cost the industry $10b/year initially to roll out 450mm wafers (paying back development only, plus the costs of building and equipping the fabs).
    This all has to be paid back out of the cost savings from bigger wafers before you can sell the fabled cheaper chips with 125% more per wafer.
    If you use the rule-of-thumb that going up a wafer size knocks 30% off the chip cost (I think this was the case for 200mm to 300mm) then a $30b/year industry (all in 450mm) will pay back the development costs back in 4 years.
    The total semi industry is around $300b/year but I don’t know how much of that is leading-edge technology that will go to 450mm, or how fast this can all be done.
    However you look at it the investment levels are enormous; even if they do make sense in the long run (which they do), nobody’s going to make money out of 450mm for several years by the time they’ve paid for the (very) expensive 450mm equipment as well as the development costs.
    Given the accountants who run most companies nowadays and the thirst for short-term returns (i.e. less than 4 years) at the expense of long-term ones, moving to 450mm isn’t going to be easy for most of the industry…

  11. Mr T.
    November 15, 2011 11:34

    Going from 300mm to 450mm gives and increase of about 225% in area per wafer. Almost twice the number of chips per wafer.

  12. greg
    November 15, 2011 01:09

    Considering investment levels…and if 450mm is the end game that alot of people will be operating on in 10 years time, perhaps you’d be better off not being a first mover?. Perhaps wait 10 years and pick up a bankrupted failure for 10c on the dollar?. Just ring 300mm for all its worth in the meantime. Even if you did get first mover advantage, your investment levels are so high, profits may be years away.

Share your knowledge - Leave a comment