NXP, TI, STMicro and the end of process development

With NXP and STMicro pulling out of basic chip process development at Crolles, and with Texas Instruments deciding to abandon basic process development, the semiconductor industry is taking a dangerous route.

Essentially the semiconductor indsutry is putting itself into the hands of the foundries. That’s OK for the time being. The foundries do a good job. But what happens when the foundries know the big suppliers are totally dependent on them? Will prices go up? Will terms get harsher? You bet they will. What looks like a saving in overhead today for the big companies could come back and hit them between the eyes in higher prices tomorrow. And how much is this saving? TI’s CFO Kevin March estimates that it will save TI between $140m and $150m a year by not doing its own basic process development. That’s peanuts to a $10bn a year revenue company. And for this mess of pottage TI is sacrificing its long-term independence. For ST and NXP, the bargain looks even more Faustian. They were sharing the cost of basic chip process development three ways. Even without taking into account the EU and national governments’ subsidy of Crolles, that works out at only $50m a year each. And ST is a $9bn+ revenue company, and NXP is a $6bn+ revenue company. For a small contribution to the bottom line in 2008, they are prepared to sacrifice their long-term independence. The end of Crolles as a centre for advanced basic chip process development marks the end of a 20 year-long strategy to give Europe independence in microelectronics technology. The strategy started with the Philips-Siemens Megaproject, continued with JESSI, and was carried on by MEDEA and MEDEA+. The argument was always that Europe needs controlled access to advanced microelectronics technology if its industries are to remain competitive. Now, apparently it doesn’t. The demise of Crolles won’t have an effect on competitiveness next year, or even in 2009 or 2010, but further out the effects may be severe. “If every company goes to foundries, using the same processes and cell libraries as everyone else, what differentiation can a company offer customers?”, asks Malcolm Penn, CEO of analysts Future Horizons, “five years down the road, when ST asks: ‘Why haven’t we got any customers’, the reply will be: ‘We can get this from loads of other people cheaper, now that they’re making them in the Congo



  1. Philippe, I hope you’re right. Maybe what Europe needs to do is to keep the process development R&D going at Crolles and turn its fabs over to foundry – open to all

  2. Philippe, I hope you’re right. Maybe what Europe needs to do is to keep the process development R&D going at Crolles and turn its fabs over to foundry – open to all.
    That will be a business model to help all of Europe’s chip companiers, not just two big ones.

  3. you are right. We have to bear in mind that there is a strange historical link between the 20th century (steel) and the 21th century (IC): foundries!
    In both case, it was / is the key for industry (and a key for world wide suprematy).
    So, we cannot look at it only as a short term business issue. The worthiness of Crolles 2 is to develop DSM ability for European industry with massive support of regional, French and EU community funds. If it stops, where is the pay back?
    We will have to pay more attention on the induced business models. Cost effectiveness as to be evaluated more at a global level than at company level as soon as it involves public founds. We do not know the right answer but it is not only a question for manufacturer’s CEOs!

  4. Peter,
    You may well be right at the moment but the industry’s history suggests that, long term, we’ll always find new uses for more transistors.
    As for wearables, sorry, I haven’t a clue
    Best wishes

  5. David – you know it is interesting to speculate what would be the effect if silicon process technology were effectively frozen at what we’ve got today.
    The benefit of new processes is that you get more highly integrated, more power-efficient and faster digital chips. Is that what we need?
    Some of your other postings indicate that consumer electronics is already overburdened with features. I don’t have access to figures, but I believe that the digital chips in a mobile are no longer the prime power users – the display is. I’d argue that simple evolution of today’s products doesn’t necessarily require process innovation.
    What we would loose are the completely new classes of products that might have been enabled, but that we don’t know about yet. Is wearable electronics feasible with today’s chips? I don’t know.

  6. You’re right Peter, there will always be competiton among foundries at the trailing edge, but how many foundries will be able to maintain leading esdge capability at 45nm and 32nm?
    IBM, TSMC, UMC definitely, Chartered maybe. SMIC unlikely.
    And that’s about it. Not much competiton there.

  7. Excellent observation!
    I quit the Crolles 2 Alliance about 1 year ago because I saw this was coming but it is a sad development for the european semicon industry. The decisions taken by these CEOs have an enormous impact because once you decide to to stop CMOS development you can never get back in the race again.

  8. I guess it depends on where you see the innovation happening.
    There are plenty of fabless companies out there doing clever stuff without process differentiation. Often on quite old processes too. And analog seems to be growing – this doesn’t need such advanced processes.
    The alternative vision of the Foundry Business is that the competition between them will remain sufficient to keep prices and terms realistic – especially if there are new entrants from say China. Is it possible that some of the companies mentioned in your article will sell their fabs to them – as has happened in the CEM business. There is demand for manufacturing on trailing edge processes.

Leave a Reply

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