You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Have you seen any evidence to support the fab build out necessary to support the staggering numbers that have been bandied about for IoT deployments by 2020?
Needed are semiconductors of all types, fiber optic cabling, telecommunications infrastructure, etc, etc, etc. Plus software...
It would seem that a mobilization much like we engaged in at the start of WW2 would be needed. I can see the graphic showing mines feeding into smoke belching factories turning out IoT's. Legions of workmen building, tunneling, laying cables...
For a more curated view of the IoT, please have a look at my twitter feed and search for "@Chuck_Petras #IoT". My interest is in the electrical power system, specifically the smart grid which I view as a subset of what is now popularly call the Internet of Things.
There is already an incredible amount of 28nm capacity in place for the mobile SoC craze. TSMC ships more than 1.3M 28nm wafers annually and that will increase by 20% this year. The transition to FinFETs is expected to start in 2015 which will free up an amazing amount of low cost 28nm capacity. 28nm also has the strongest design ecosystem with more than 100 partners including 39 vendors offering more than 6,000 pieces of IP. GF, UMC, and SMIC are also adding 28nm capacity.
IoT chips are very small in comparison to SoCs so looking at the grand scheme of things I do not see any wafer shortages coming at 28nm and above as a result of IoT. Sound reasonable?
Don't forget to add mems to the list. This industry alone will be as big as the computational and communications semiconductor industry. Much of this need will be met by using larger node fabs that were considered obsolete and now will have there economic life span extended very dramatically. This will change the economics of the semiconductor industry more than anything since it's founding. Did anyone else catch the statement by Morris Chang about extending leading edge technology backward towards older fabs? It's interesting that TSM has become one of the top builders of mems in the space of just a few years. This is going to change the dynamics and economics of not only the semi industry, but medical and many other mega markets. Performing many functions at the nanotech level by on a massive scale has the possibility and probability of massively changing everything. Desalinization and Solar (mems controlled mirror and lenses?) are but two huge industries that may come into their own from this.
peter gasperini
I suspect the fabs are going to HAVE to lag such demand - it would be suicidal to build out capacity and then wait for one or more consumer electronics companies to find a design/configuration that has enough of a value proposition to coax discretionary cash out of the pockets of cash-strapped consumers.
The transition to mems will be far faster than the standard semi industry adoption for much of the knowledge and equipment is mostly transferable and it's just the incremental knowledge we have to gain. Even much of this will just be massively scaling down processes already in use at a staggering savings that will cause many industries and businesses to expand rapidly, just like the traditional semi industry did to many areas. Also there are far more educated, skilled people now working in a far more flexible capital structure than when the traditional semi industry came into being. This doesn't even take into account the massive power of the net that wasn't even there when the semi industry started. We are going to be looking at one of the largest technological revolutions that will also happen at a revolutionary speed. It will only be limited by the human imagination, which even that has vastly expanded world wide. Hang on tight, it will be a wild ride.
The fab needed for IoT devices will not be cutting edge high capital intensive fabs. It will be targeted for MEMS and analog fabrication mainly. I do not think any IoT device (except processor) will use geometry below 65 nm to fabricate chip
Barun, I disagree with you on this. It may start out that way, but I feel mems will shrink substantially just like semis and the technology will advance just as fast as semis. This will allow complex, multi stage functions on a single chip. Think of a chemical or bio lab that takes up a whole room put on a single chip. Just like Intel vastly underestimated the demand for processors initially, I think the same is true of mems. Right now we see only the very tip of the number of functions we will see on a mem. I also believe mems will become like fpgas to be able to programmed to be programmed to a number of complex functions on a single chip. A super complete blood test would be but one of thousands of examples. To think mems won't advance in technology, node, flexibility, technology while dropping in cost like semiconductors is to not learn the lessons of the geometric rate we are increasing our knowledge across the scientific and applications fields right now. The mems plants will end up being as every bit as complex as our semi plants are, just in different ways. Solar is basically a mem that processes light and they are advancing at a rapid clip going from single junction to triple junction and getting over 40% efficiency. Soon I see CSV solar with built in mems focusing so you will have a flat panel that focuses towards the sun on a microscopic level, sort of like a DLP controlled light, but this will be lenses instead. Much of this is speculation, but look at what's happening in many fields that we could barely imagine just twenty years ago, except in a science fiction novel. The pace of change in everything is vastly picking up speed and technology is moving far faster now than twenty years ago. This is why the mems revolution will take place far faster than the semi revolution for it's taking place in a time when technology is moving at least four times the speed due to increased computer power, number of educated people, collaboration from the net and far cheaper and better instrumentation. I would appreciate your thoughts, views and comments on this, because I deeply value and appreciate people who can help me improve my logic and reasoning. I value people who help me learn and understand better the world we live in. Thanks, ARTHUR W. HANSON
Barun, I disagree with you on this. It may start out that way, but I feel mems will shrink substantially just like semis and the technology will advance just as fast as semis. This will allow complex, multi stage functions on a single chip.
To think mems won't advance in technology, node, flexibility, technology while dropping in cost like semiconductors is to not learn the lessons of the geometric rate we are increasing our knowledge across the scientific and applications fields right now.
Actually I have not meant that mems will not advance in technology. Mems is one the next revolution in semiconductor industry and the technology will become more complex and more useful. And as you have very rightly pointed out lab-on-chip is one the revolutionary application of mems. PCR testing on lab-on-chip will demand a FPGA like solution. Currently universities are researching on development of FPGA like lab-on-chip which comprises of more than 20,000 nodes. And I expect this technology, once matured enough, will have much faster growth than semi growth. The main reason is the application is in healthcare area which has far more impact and demand on human society than an electronics gadget.
What I want to mean setting up mems manufacturing units or doing a design for mems chip does not need heavy investment like SoC design and manufacturing at least till now and it will be remain true for few more years. In fact this is a boon for mems research.
If you are interested I will recommend to read books, publication of Krishnendu Chakrabarty of Duke University. In his several papers he has explained how an FPGA like device will be used in future lab-on-chip and how design automation can help user to implement very complex PCR functionality on such chip very easily
Barun, I am neither an engineer or financial person, I am a compulsive reader, although I have been a tech in the past at LLL. I am self trained, although I have taken numerous short courses and taken classes in chemistry, calculus, thermal systems and physics way in the past. I do watch trends and trade on them for a living. I connect dots and follow money trails. I seriously think you might be underestimating the technology that will be going into mems, especially in handling light and its many uses. Amat and Siemens have backed Semprius and I think if a mems could focus the light on a flat fixed panel this could be a killer app that would shake up the worlds energy picture. Combine that with this new battery and the whole energy world could change radically. I thank you for your insights and as a sounding board. ART
