Neelie Saves Europe From The Brink

It looks as though Neelie Kroes’ 10/100/20 programme to invest $100 billion to get Europe a 20% share of the chip market has come none too soon – Europe’s share of global semi capex this year will be just 3%, says IC Insights.IC Insights forecasts capex spending by Europe-based ST, Infineon, and NXP and all other European semiconductor suppliers combined will amount to less than $1.5 btoo soon illion in 2013.

Asia-Pacific is the capex king – accounting for 53% of capex marketshare in 1H13, down slightly from its 55% peak in 2010, says IC Insights.

The capex was spent mainly on fabs to produce DRAM and flash memor and microcontrollers, and to bolster wafer foundry operations.

Mostly on account of spending by Intel, GlobalFoundries, Micron, and SanDisk, North America accounted for 37% of capital spending in 1H13, a few points higher than the steady 29%-33% share it has held since 1990.

Led by Samsung, Intel, and TSMC, there are nine semiconductor suppliers that are forecast to spend more money on their own than Europe will spend collectively in 2013.

In IC Insights’ opinion, IC manufacturers that are currently spending less than $1.0 billion a year on capital outlays will find it just about impossible to continue being able to manufacture using leading-edge digital processing technology.

Which would be bad news for the future of FD-SOI – had it not been for the support of Neelie Kroes.



  1. Ok Stooriefit, I understand your point on the processor architecture now. And actually what I said was the best ARM processor would be one made on Intel’s process – as you seem to agree they have a massive lead in semiconductor processing. But given that won’t happen can ARM on TSMC compete with Quark on Intel ?

    Moving back to FD-SOI, I’ve always said it’s a great idea at 28nm. It give very good performance at reasonable cost, albeit still massively more expensive than 28nm planar – more expensive wafers and more processing steps to put the back gate in. Likewise it should be ok at 20nm, what they are calling 14nm for marketing reasons. However the problems start to occur in the process after that, the real 14nm one which ST are calling 11nm, where things break and break badly, ST themselves admit there is a problem but think they can make it work and won’t need FinFET until their “8nm” (in reality 11nm) but I’ve never found anybody outside of ST who doesn’t just look skywards.

    In any case by this point you’ll probably need equipment that is only available in 450mm version so they’ll have to build a new fab or get out of the game anyway. But I don’t see building another large fab in Europe as the non-starter you imply. Indeed it is a unstated prerequisite of 10/100/20 to build several ! Intel and GlobalFoundries are investing heavily in Dublin and Dresden proving the economics do stack up once you remove the red tape and give the right incentives. The money spent developing FD-SOI could have contributed towards building such a fab using a process directly from IMEC.

    Regarding IMEC’s subsidy that is rather different. IMEC has to make money on its CMOS programme, but does get a subsidy to carry out socially aware projects such as heatlthcare and also more blue sky work. FD-SOI conversely has notr made a Euro yet, and without ST-Ericsson there are no projects large enough to justify its on-going development.

  2. That’s the way I think too, Stooriefit, I think FD-SOI is a welcome move to get process advantage, it’s welcome as an alternative to finfet and it’s welcome as a pioneering European initiative.

  3. My clumsily made point was:

    Nearly a decade ago AMD were wiping the floor with Intel on desktop performance. Intel could have just licensed another processor architecture and cleared AMD off the map with their process advantage (I think it was you said the ideal processor would be ARM on Intel’s process in an earlier comment), but they didn’t – they developed CORE (a 21st C architecture) and got back in the x86 game.

    ST might be able to pull the same trick in the process game by having unique FD-SOI capability. Everyone is swping the floor with ST on process, but FD-SOI has a chance of reversing that. The trouble with licensing the IMEC finfet process is that everyone else can too, and as we have seen with memory, it isn’t economic to do fab in europe if the same technology is available to do in asia, so process differentiation is the only way to keep your chip baking in Europe. You are also ignoring the extras FD-SOI may allow in SOC (being much closer to standard planar process, but with speed & power advantages) that the IMEC’s finfet would need further and more complex development to achieve.

    There has been much bitching in these comment pages about semicon firms no longer being prepared to bet the farm on the next process, and now one does they get bitched at for doing it.

    BTW you confused me with your subsidy prod – IMEC is ~18% funded by Flemish and Dutch governments.

  4. Not sure why you’ve brought x86 into the discussion but Haswell is NOT a vintage processor design and x86 vs ARM are now delivering similar performance apart from at the very high performance end where ARM architecture cannot compete due to its fundamental limitations, and at very lower power where the Intel architecture obviously cannot compete with things like the M series.

    However IMEC FinFET vs FD-SOI is a totally different game. One works well and one is an expensive tax-payer supported technological dead end. ST and IBM are the only semiconductor companies not to licence the IMEC work and I’m sticking to my statements that this is a fiundamental mistake by both companies.

  5. Mike – the counter factual is that we’d all be running computers based on ARM or MIPS if Intel had done the sensible thing and licensed a decent processor architecture, rather than investing all that effort in re-working the micro-code of x86 and x64 in the CORE programme. Maybe a sensible thing, but we’d be properly sunk if they had because there would only be a single microprocessor manufacturer on the planet if you combine Intel’s process lead with a modern processor design reather than vintage one…
    What keeps us western R&D wallah’s in business is that the marketing guys are always looking for differentiation, and the (grossly simplifying) model of eastern “me too” commoditised products just don’t have the margins. You don’t get the potential for differentiation by taking out non-exclusive licenses, and exclusive license costs are of the same order of magnitude as the R&D costs anyway.
    I imagine ST first saw Neelie’s plan when it was drawn in the condensation of her bathroom mirror, so they’ve kept their capex in their back pocket until it can be spent under the auspices of her programme, and therefore release the EU cash. It makes sense – spend your cash only when you can gear it against some of the tax payer’s. (Struggled with where to put the apostrophe there, but decided I’m probably the only tax payer in the EU – that’s what it feels like anyway!).

  6. If ST et al hadn’t spent so much developing FD-SOI and instead licenced working technology from IMEC they’d have had the money to upgrade their fabs earlier.

Leave a Reply

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