Samsung Joins The Litho Party

Samsung has followed Intel and TSMC in buying a stake in ASML and agreeing to contribute towards its R&D costs for 450mm and EUV.

Samsung will buy a 3% stake in ASML for Euros 503m and will put up Euros 276m for ASML’s R&D programmes.

Intel paid $3bn for a 15% stake in ASML and said it would put another $1bn into the R&D projects.

ASML has said it is only offering a limit of a 25% stake to its three top customers.

TSMC is paying Euros 838m for a 5% stake in ASML and another Euros 276m for the R&D projects.

ASML says it expects 450mm tools to be available in 2018, but TSMC has said it expects to be using 450mm tools in 2014.

One programme, the EUV programme, needs to increase its 4000 engineers by 30%, says ASML.

ASML says it has now completed the R&D funding programme and doesn’t plan to ask other customers to participate.

The investments by Samsung TSMC and Intel mean that ASML is now the monopoly supplier of leading edge lithography tools.

Japanese rivals Canon and Nikon are already behind ASML and the $1.38bn being put into ASML means ASML will increase its lead over the other companies.

The programme also gives Intel, TSMC and Samsung a pretty much unassailable lead in process technology because they’ll be guaranteed first deliveries of the latest and most advanced chip manufacturing equipment.

ASML says that the Euros 3.85bn billion raised from selling an aggregated 23% non-voting stake to its major customers will be returned to existing shareholders through a synthetic buy-back.



  1. Interesting Mike, thanks. What a frightening consolidation of power, though.

  2. I think you can count Nikon and Canon at least down if not out. All of their top potential customers are now in the ASML camp with R&D contributions and shareholdings to match so it would take something far better than ASML’s offerings to tempt them back. This only leaves tier 2 and 3 players, all of which together don’t spend enough capex to make the market worth entering.
    The Japanese ministry responsible has tried to kickstart and EUV program several times based upon the EC model but with little success.

  3. I hope you’re right, anon, having no competiton is unhealthy. But the Japanerse are showing little sign of getting their act together on this one. And I think the Japanese government may be getting a little fed up with putting money behind consolidations which don’t work.

  4. i wouldn’t count out Nikon and canon so fast, these are also companies with billions

  5. They might have if Hynix and Toshiba had bought all 25% of the company originally on offer ๐Ÿ™‚

  6. Good point, didn’t think of that. Also the reason why they only took 3% rather than the expected/offered 5% is unclear to me.
    Then again the difference in benefit for Samsung will be minor: if you want to sell semi manufacturing equipment you do not put Samsung on the second row no matter how much/little they care to invest in you.

  7. I wonder if that is why Samsung left it so late in announcing its decision, Martijn, or am I being unduly cynical?

  8. @ david
    true there is 2% left but for share holder regulations the discussion needs to be closed this month (tomorrow). There is 0% chance GF will close a deal in these 2 days.

  9. Yes indeed martijn but originally ASML said it would sell up to 25% and Intel took 15%, TSMC 5% and Samsung 3%. So, theoretically, there’s still 2% which GF could go for.

  10. Well actually he did say he was going to see IBM about it quite soon, anonymous.

  11. Thanks David & Mike – it is interesting to see how this stuff works in practice.
    As for the high level, I suppose the only guys who have a few bn lying around in loose change are Apple, Intel and one or two others… As has been said above, there are no guarantees that this cash buys timely capability that delivers on its promises for future economies in actual products.

  12. Is there much more of a high level than the CTO of a top 10 semi IDM ?

  13. IBM is not a significant player in actual manufacturing, so they don’t need 450 mm.
    For process development they need EUV of course, but the difference between the current generation of EUV and the next (or actually the one even beyond that) generation about which this investment is, is mostly for manufacturing and not core “middle of the wafer 1 die per hour” resolution.
    So IBM can keep doing what they are doing without getting hurt by not investing.
    For the alliance, i.e. the partners using their technology it’s a different story. Great technology without a proper tool to manufacture it on is not going to fly, so I would say they “missing” party is GF, not IBM.
    Meanwhile ASML does not want to partner with just any old institution, so only if the top 3 said no would others have been approached, and they didn’t.

  14. Funnily enough, Stoorieft, I was with ST’s CTO Jean-Marc Chery this morning and so I asked him your question. His reply was that this was something that would have to be decided at a high level. The same applies to the college of nanotechnology at Albany which had one of the first two EUV machines. Jean-Marc thought it was more serious for GloFo.

  15. @Martijn : One of Future Horizons’ recommendations to the EC was to build a European 450mm fab for non-leading edge nodes before someone else did. Eat lunch or be lunch !
    Our report also highlighted the likely consolidation in the industry and the need to make sure Europe has a world class player afterwards either through consolidation or fab-sharing.
    @Storriefit : Samsung is still in the IBM alliance so maybe IBM will use them to fab devices. In the longer term I still believe IBM will sell its semiconductor research arm to GF though and concentrate on what it does best. It cannot hope to match the R&D spending of Intel, TSMC and Samsung, let alone the capital expenditures necessary to keep at the bleeding edge.

  16. I’d appreciate it if one of you chaps who know the business could fill me in on where this leaves IBM & the Common Platform Alliance. Or have they another way in?
    I find it hard to believe that IBM would be prepared to be left behind if it meant they became properly uncompetitive.
    Does it actually mean they become uncompetitive in the commodity silicon market, not the really high end, where their Z196 sits for example? This is nowhere near the cutting edge on feature size but is no commercial weakling…
    Maybe Intel and Samsung are moving down the food chain, digging like meerkats to catch beetles. The smart money could be still lying under the acacias batting flies with its tail waiting for the nice juicy ruminants to appear.

  17. 450 mm is do-able, question is will it be meeting requirements for latest nodes (N14/N10 by the time it’s on the market).
    Or, will the big 3 be willing to build a 450 mm fab for non-leading nodes, just to outprice the smaller guys?
    Currently the smaller guys stay alive by picking up the slack on non-leading nodes, if the big 3 can undercut them by 20-30% due to 450mm production massive consolidation has to follow.
    That will leave them with plenty of time (to market) to play with EUV on leading nodes (on 300 mm)…

  18. I meant EUV works – it’s just the processing speed is too slow for commercial application at the current time but it’s the only option for the post CMOS era.
    450mm is also the only game in town and it’s these negative comments that Europeans are so good at making that will condemn us to the Dark Ages. Of course it will take time to get it right, 300mm did as well, but once up and running those shiny 300mm fabs are going to have cost structures so uncompetitive it will hurt. We need another Pistorio to ‘make it happen’ here, although of course not investing in ASML has already condemned GF, UMC, Japanese and European companies to a much later position in the order book.

  19. I think you’re right, SecretEuroPatentMan, these technologies could end up like the Fifth Gen Computer Project or bubble memories as white elephants. There are few things more compelling than a delusion whose time has come.

  20. ‘just a bit slow’? I think that’s the understatement of the year. And in my experience, if a technology guy tells you soemthing will be working in 2018 – i.e. 6 years hence – it means it’s very dodgy whether it’s going to work at all. I think EuroSecret PatentAgent Man has a point. Intel, TSMC and Samsung could be chucking their money away on a chimera – but they have plenty of money to chuck away.

  21. SecretEuroPatentAgentMan

    Considering how much gnashing of teeth we hear for the next steps in size and geometry I am surprised how willing these companies are willing to cough up serious money for such a high risk project.

  22. EUV works … it’s just a bit slow at the moment ๐Ÿ™
    And I think ASML are showing just how hard a 450mm table is by saying 2018, not some seat of the pants date their customers would like.

  23. Kiss of death for GloFo then? Assuming, of course, ASML can get 450mm and EUV to work.

  24. Alas poor GlobalFoundries, we knew them (quite) well. ๐Ÿ™‚

Leave a Reply

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