Electronics Weekly Electronics Design & Components Tech News
The SIA held a webinar earlier in the week which resurrected the arguments of yesteryear – e.g. foreign governments spending much more than the US government, cheaper labour in foreign countries, need for a level playing field etc etc.
It missed the key problems: Who in the US can fill a fab? Which US chip companies want US fab? Is the IDM model dead? Could companies share a fab? If so, is there any point building a fab operating on the foundry business model to compete with the giga-fabs of Samsung and TSMC?
The Japanese tried the joint fab approach 14 years ago with the Advanced Process Semiconductor Foundry Planning Company (APSFPC) but failed because different companies wanted different tweaks to the process.
Even then competing with TSMC was seen as impossible for any single company. “Japan has 11 IDMs at the moment but their fab capacity is not big and they can’t compete with TSMC which has 60,000 wpm fabs,” said Hirokazu Hashimoto CEO of the APSFPC in 2006.
The situation is even more forbidding today with TSMC using 100k wpm fabs with a clear lead in process technology.
If joint fabs are not the answer, then which companies in the US could afford to run one of their own?
The only US companies that could fill a fab are Qualcomm, Apple, Broadcom, Intel, TI and Micron.
Qualcomm, Broadcom and Apple have shown no interest in running silicon fabs. Intel has recently indicated its intention go fabless at the 3nm node for its leading edge products. TI doesn’t need leading edge fab. That leaves Micron.
Surely all this lobbying effort can’t be going in just to support Micron.
The SIA pitch is pretty crude: Give us the money.
It wants $20 billion over ten years to achieve the USA’s current 14% share of new worldwide manufacturing capacity or $50 billion over ten years to achieve a 24% share of new worldwide manufacturing capacity.
The amounts seem puny – Intel alone is budgeting $17 billion on capex just for 2020.
The SIA argues that it is 25% cheaper to build and operate a fab in Korea, Taiwan and Singapore than it costs in the US, and 50% cheaper in China.
It argues that the US has 12% of the world’s installed manufacturing base while Asia has 77%.
It argues that foreign governments pay 40-70% of the cost of building a fab and that the US government should match it.
But the arguments are meaningless if the companies don’t want to do it.
Just ask the EU. A few years ago, the EU offered the European chip companies an €80 billion support programme if they invested enough in Europe-based chip manufacturing to achieve a 20% world market share.
The European companies said they’d like the money, but they weren’t going to build fabs to achieve 20% market share.
And one expects the US companies to say exactly the same.
An excellent insight robetc – this situation may be a catalyst for major innovations in IC technology from outside the US.
Well I think the US has done the correct thing in letting the evolution of the industry take its natural course with advanced logic foundry going to Taiwan and Korea which are solid friends with the US, robetc. Throwing dollars at the US industry to try and perpetuate the IDM model is ridiculous. At the same time the US has maintained its grip on the key constituents of advanced manufacturing equipment and leading edge design software. I expect no ine could keep an advanced fab running for long without the help of the equipment companies’ engineers.
I wouldn’t be complacent though. The world is turning fast and technology, like water, has a way of finding its own level. Expertise is a mobile and convertible commodity, and there is lots of both expertise and money floating around the Far East. I imagine that a lot depends on how the US manages bleeding edge new technologies during the next decade. The more the US steps on the toes of its competitors, the more determined they will be to develop their own resources. It will be a fascinating period, technically, commercially and politically, for process development.
Hoisted by our own petard. TSMC, Chartered and the like exist because we want better process at rock bottom prices. Eventually the standard and cost of living will rise for the places in which their foundries exist. Prices will rise and in Europe and the USA there will be little in the way of indigenous talent to turn the ship around. R & D has a habit of geographically following production. Although the US is still at the heart of process technoology there is no reason to suppose that will remain the case. I mean, the very idea that the Japanese could build better cars and motorbikes than the British. Ha.
It points strongly at the fallacious idea that you can easily disconnect a connected world, and that forriners somehow haven’t got what it takes. Trumpists and Brexiters take note.
Actually robetc the US isn’t at the heart of process technology…it all comes from imec in Belgium!! And what a brilliant job they’ve done 🙂 That is something to shout home about.
A sad indictment to the Boston Consulting Group who clearly missed or ignored the points you listed; they should be holding their heads in shame.
Also there’s no reason why a jointly-owned fab shouldn’t work … they key is NOT to run it as a joint venture (and suffer the endless wrangling over self-interests Hasimoto-san pointed out) but as a FOUNDRY under an operations-driven CEO with strict process rules and lead-times … just like TSMC.
They is also one other key missing point; if all these fabs are to be built in the US, they will presumably all be directed at logic not memory which means a strategy to take business away from the one firm the USA (and the rest of the world for that matter) is wholly dependent on. Even if its logic and memory, that means a strategy to eat Korea and Taiwan’s lunch. Nice one Uncle Sam. With friends like these who needs enemies?
And as you point out … the noise from the US chip firms is deafening (not!); they (and Wall Street alike) clearly can’t wait to dump their fabless SC model to return to an asset-heavy, capital intensive IDM strategy with the consequent hit to their bottom line, P&Ls, stock price and bonuses.
There has not been a single US chip firm that chose to build a fab offshore vs the USA because of the better /cheaper costs and/or un-level playing field on offer elsewhere!
Wise words, Malcolm, it seems to me the IDM model is dead – replaced by the fabless-foundry model – trying to go resurrect IDMs is a hopeless strategy except for analogue. For the SIA to go along with this idea is quite extraordinary so my feeling is that, like the European chip companies, the US industry can smell the government money and don’t want to turn it down, but have no intention of using it to build fabs.
Indeed … hence why now the SIA is headquartered in Washington DC and stuffed full of suits vs Silicon Valley where it was founded and driven by industry guru’s. Hey ho…fastest way to the dole queue!!
Yes indeed Malcolm, I can’t imagine Wilf Corrigan, Bob Noyce, Jerry Sanders, Charlie Sporck and John Welty thinking that putting the SIA in Washington would be a good idea.
Well Yes Duncan the whole world relies on Taiwan-made semiconductors and I think it’s that which worries the Americans
I thought the only thing that kept Micron going at one point was the US Governments decision that semiconductor memory was a strategic industry / resource.
Last week there was talk of the USA returning to the moon. I wonder what percentage of Apollo 11 was foreign made; Probably nothing. I wonder if the best subsidy the US Government could provide would be to say that Government contracts – Space, Defense, Supercomputers, and even POTUS’s phone (which he relies on so much) had to use Semiconductors made in the USA. They wouldn’t subcontract building the latest Military Hardware to Taiwan but it probably relies heavily on Taiwan made semiconductors.