There were two complementary pieces of news last week – that ST had signed Samsung as a foundry for FD-SOI and that ST dropped out of the top ten.To succeed in the the chip industry you need to be brave. ST used to be a brave company – investing in MPEG2 and MEMS years ahead of the market, fighting its way into the cut-throat memory market, competing with the best in the world for mobile markets.

But somewhere along the line ST stopped being brave. The most obvious sign was not equipping the Catania fab – from that moment on ST was never going to have the capacity to grow rapidly.

ST-Ericsson died for not having leading edge capacity. The memory business was shuffled off to Micron. ST turned in on itself retreating to safe, profitable, slow-growth, behind-the-leading-edge product areas.

It’s a respectable strategy, many people in Europe think it’s the right strategy for European companies, it’s a strategy that the financial community support, but it’s a strategy which means the companies which follow it inevitably slip down the league table of leading companies, unable to generate the funds to invest in new generations of technology – becoming safe, predictable and timid.

Some would say that safe predictable and timid is OK for Europe. Others say that Europe should do better than that with its distinguished scientific history and its intellectual strengths.

If ST had gone ahead and equipped Catania it would have had enough capacity of its own to bring FD-SOI to market and start generating growth and wealth to re-invest in the future. It didn’t, so now ST must seek capacity elsewhere.

That’s not ideal but it’s OK. If the market recognises the benefits of FD-SOI then ST will start to grow again and Europe will have a technology champion able to compete with the best in the world.

Such a strategy needs brave leaders and, luckily, Europe has a brave politician in Neelie Kroes and a brave technologist in Jean-Marc Chery.



  1. You’re absolutely right about Crolles. When Crolles 1 opened it was obviously not a lab but a production fab, but the French have been happy to pretend it’s a lab for 23 years for the purpose of subsidies. I would have thought that the land which spawned Machiavelli could devise a similar fiction. And, of course, Crolles is making complex SOCs using IPs from all over the place and how good is Intel at making those?

  2. Catania was a real fab, not a toy, and hence could NOT be subsidised. And as ST France have found out the recent expansion of Crolles has made subsidies there on very dubious ground.

    As for competing with TSMC, I reiterate that Chery’s costs from Crolles include this subsidy and if the EU reject it then those costs will go through the roof. Also if anyone at TSMC Fab 12 was turning out 70% yield Morris would have their head on a plate !!! Even Intel won’t ship anything from wafers that low.

  3. Well if the French can find a way to subsidise a lab/fab i would have thought the Italians could have found a way to do it too. Chery says Crolles is very attuned to TSMC cost/pricing. “The challenge is always how to align the wafer cost to TSMC’s selling price,” he says. Yield matters and Chery says that, at Crolles, “the number of good circuits per wafer is between 70% and 90% depending on complexity and ramp up.”

  4. It’s illegal to subsidise manufacturing so we know M6 would definitely have not had any after the initial capital investment which is allowed.

    Crolles is classified as a Lab/Fab and so skirts around the gray area of the rules, although as ST are finding out the EU is now taking a much closer look at just what is being subsidised there.

    In any case if you are relying on subsidies to compete with TSMC then really you are not as price competitive as you say Chery claimed.

  5. Well 5k wpw is the capacity, Mike, I don’t know how many actual wafers are being run every week but would assume it’s well over 4k. And with my accustomed massive respect I would point out that we’ll never know if M6 would have had subsidies on its manufacturing because it never produced a wafer, though I concede it’s unlikely that they would have been French subsidies. M6 certainly had about half a billion euros in Italian subsidies negotiated by Pistorio with local and federal government to get it off the ground.

  6. Crolles is only competitive on price because of French subsidies. Catania would not have had these. 5k wpw is ‘noise’ and there is no way you can compete with TSMC at these volumes

  7. ….and Jean-Marc Chery should thank the sicilian IT engineers, hired for the M6 Catania project, for setting up and deploying the fab automation system in Crolles. This remark just to stay on the wave of “where ST would be if it only had the balls of fabbing up M6 in Catania”, a plant supposed to be twice the size of Crolles’ and therefore with much less operating expenses..

  8. Well I think you may do ST an injustice there Mike, they run their fabs very lean – particularly the Singapore fabs. And I know what you’re next remark will be, it’s ‘How can a 5K wpw 300mm fab at Crolles compete on cost with a 100k wpm 300mm fab at TSMC? I asked Jean-Marc Chery that question and his reply was: “We are competitive in terms of purchasing price. At 5K wpw, below 40nm the advantages of the dimension of scale is getting lower and full automation means we do not need the high volume to be competitive; with a high level of automation we can manage average volume with strong efficiency.”

  9. That is undoubtedly true, Jack, the digital business was left underinvested. The 2005-14 capex-to-sales ratio says it all. When Bozo took over in 2005 it was 26%. By 2007 it was down to 11.4%. Last year it was 6.6% and in Q1 2014 it was 6%. You can’t grow a digital semiconductor business that way, no way Jose.

  10. ST-E was on a fatal decline long before FD-SOI was even in pilot production. They did eventually ship samples of one device before they died but everyone could see they were dead and couldn’t commit to it.

    Anyway rule 1 of our industry – customers do NOT buy processes, they DO buy applications. ST-E did not have the correct desgns available that Qualcomm offered despite having access to the same TSMC 28nm process, Without these no customer needed them.

    And even if FD-SOI had come earlier at a Catania fab, TSMC would still have had a cost advantage over any other process simply because it’s TSMC and far far better run !

    What will be interesting to see is whether TSMC needs to offer special ‘pricing points’ to compete with Samsung now, not that Samsung has ever been interested in customers smaller than say Broadcom anyway.

  11. Main reason STE died is they are 2 generations behind Qualcomm and Broadcom. STE is always the first one to announce such as licensing Eagle(A15), but in the end there is no CPU out from it. Also, Bozo was not fond of pushing Digital biz out to the front. Investment was not in priority compared to the SPA (Sensor, Power, Auto) group. As such, it bled slowly to death.

  12. But not, Mike, to a leading edge volume FD-SOI process. A full-scale, leading edge, 28nm FD-SOI line in Catania could have secured success for ST-E. The lack of it, at the right time, meant that ST-E only had all this ‘legacy’ (i.e. obsolescent) 40nm and 65nm product to sell which is what is blamed for bringing them down.

  13. “ST-Ericsson died for not having leading edge capacity”

    I don’t think this was the case at all. ST-E had excellent access to all of TSMC’s 28nm technologies and I believe was the one of the first to attempt to use TSMC’s HKMG process on a risk basis.

    ST-E died for not having customers – no other reason !!

Leave a Reply

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