Which way the semiconductor industry is going?

[shaz]

Hi all,

A very happy new year to all.

I have technical background of the semiconductors (design side) but very weak industry/business knowledge and I am doing an MBA project at Cranfield School of Management with a focus on:

"In coming years, how the industry will transition to a new (bulk CMOS silicon or other) technology, and how today's players devise strategies to prosper (keeping chips cost down) through the transition"

I would like everyone who know the semiconductor industry to express their views in light of the picture (below) and the MBA project focus.

In the quest for info, I came across ITRS (International Technology Roadmap for Semiconductor Industry) website and got a nice picture re: "Emerging Research Devices" from:

http://www.itrs.net/Links/2011ITRS/2011Chapters/2011ExecSum.pdf

View attachment 5725
Clearly, this picture shows three pathways. So, will the players "niche out", will they do everything (IDMs), will some wait and be washed away, will the value chain get more fragmented etc.

Please comment and thanks in advance for your time.

Regards,
Shaz

<script src="//platform.linkedin.com/in.js" type="text/javascript"></script>
<script type="IN/Share" data-counter="right"></script>

 

[Michael Clayton]

Your location says Cambridge, UK....nice town, I was there last year for a start-up, and enjoyed pub-crawling along the river Cam with the engineers at that startup.

An MBA for a design guy sounds like you are headed for startup activity some time in future.
And one thing you might consider, is that there is still no money in manufacturing compared to "less skin in game" design and test fabless model...but fabless model has more competition, including creative destruction of your best ideas, vulture capitalists, and lawyers, so pay attention to CONTRACT LAW while at school. [That was Bill Gates only take-away from Harvard it seems.]

The ITRS glossary has risk-reward comment I find interesting: "Cost-per-Function Manufacturing Productivity Improvement Driver—In addition to the “Moore’s Law” driver of functions/chip doubling every two years , there has been a historically-based “corollary” to the “law,” which suggests that, to be affordable and competitive, manufacturing productivity improvements must also enable the cost-per-function (micro-cents per bit or transistor) to decrease by -29% per year. Historically, when functionality doubled every 2 years, cost per function must also reduce by half every 2 years (-29%/year average). On average then cost-per-chip (packaged unit) , for affordability, could remain approximately constant (requires both flat chip cost targets and flat back-end packaging targets to remain constant), If functionality doubles only every three years, then the manufacturing cost per chip (packaged unit) can remain flat if the cost per function reduction rate slows to one-half every 3 years.(-21%/year average). It should be noted that this simplistic manufacturing cost affordability model, used as a first-order driver for the ITRS, does not take into account the economic supply and demand market complexity of actual external market environments." ITRS Exec Summary attached to your discussion...thanks for that.

As you can tell, this 2011 intro to possible futures is already old and future is extremely risky for IDM's and large foundries.
Europe seems to be going full bore ahead on 450 mm and new materials technology, but that may be due to state subsidies not found so far in US.
IBM may have some new tricks up its sleeve, to disrupt the assumed finfet flows for foundries.
Intel has already slowed down Fab 42 startup, perhaps to rethink design-process-integrations.
2013 will be interesting if Moore's law scaling proves too costly, and the only answer (limited to really big companies) is 450 mm wafer size. That completely destroys legacy IDM companies with DRAM or Flash or MicroProcessor dreams in my opinion....but not the Analog, SOC or Power IDM's necessarily...again in my opinion.
I see same discussion on Linked In.
You should find lots of thoughts on this topic using keywords on Google suggested by this ITRS roadmap summary.
But you may not get much help naming survivors and losers, in terms of companies or technologies.
This industry if famous for liar's poker games in technology papers vs actual early rampups.
Good luck with your project.

 

[Daniel Nenni]

The semiconductor industry is going completely fabless. Consolidation will hit the semiconductor ecosystem from all directions very hard in 2013. Even the foundries will see consolidation with excess capacity at all nodes with the exception of 28nm and below.

This is not just a cost/expense problem, this is an innovation problem. The small to medium companies in the ecosystem must innovate to survive. FinFETS is just the beginning.

On the good side social media has come to the semicodncutor ecosystem and will continue to facilitate collaboration and innovation.

All in all it will be a very interesting year for us all!

 

[Daniel Payne]

The two semiconductor technology directions that I see are FD-SOI and FinFET, at least for the next several years.

 

[shaz]

Thanks all and especially Michael, Daniel Nenni and Daniel Payne.

I decided about the topic after talking to a couple of CEOs here in Cambridge (about the topic to choose for my MBA in Strategic Management) and both were talking about the same thing: how today's players devise strategies to prosper (keeping chips cost down) through the transition" in next 5-8 years.

I like the confidence of Daniel Nenni that the Industry is going fabless. I work for an IDM and it is bizarre that even though I have joined IDM after working for a fabless company, I see no difference and it feels like I am in a fabless company as the two businesses (design and fab) are completely separate.

If we take this assumption that all companies are going fabless (in transition) and that there will be few pure-foundries and that the cost of making these chips will keep on rising, then in order for all to prosper (make money), the only way forward is innovation, collaboration and strategic alliances. This sounds straightforward but each of these is a world in itself.

I would welcome anyone who can critique this notion.

Anyone with a past knowledge - the industry made a technological transition from bipolar to CMOS, are there any lessons to be learnt from the previous transition.

Thanks again,
Shaz

 

[Daniel Payne]

Shaz,

There was another transition in-between bipolar to CMOS, it was called NMOS and it suffered from high standby power. NMOS lasted until about 1982 or so, then CMOS was rapidly adopted because it solved the standby power issue.

 

[Daniel Nenni]

Thanks all and especially Michael, Daniel Nenni and Daniel Payne.

I decided about the topic after talking to a couple of CEOs here in Cambridge (about the topic to choose for my MBA in Strategic Management) and both were talking about the same thing: how today's players devise strategies to prosper (keeping chips cost down) through the transition" in next 5-8 years.

I like the confidence of Daniel Nenni that the Industry is going fabless. I work for an IDM and it is bizarre that even though I have joined IDM after working for a fabless company, I see no difference and it feels like I am in a fabless company as the two businesses (design and fab) are completely separate.

If we take this assumption that all companies are going fabless (in transition) and that there will be few pure-foundries and that the cost of making these chips will keep on rising, then in order for all to prosper (make money), the only way forward is innovation, collaboration and strategic alliances. This sounds straightforward but each of these is a world in itself.

I would welcome anyone who can critique this notion.

Anyone with a past knowledge - the industry made a technological transition from bipolar to CMOS, are there any lessons to be learnt from the previous transition.

Thanks again,
Shaz

Hi Shaz,

It is interesting you say that current IDMs are not truly integrated with design and manufacturing. The top fabless companies that I work with operate like IDMs now. They have process technology groups, yield groups, lithography and test, etc... Seriously, they look even more like an IDM than current IDMs since all groups are tightly integrated. I'm talking about Qualcomm, Broadcom, Xilinx, NVIDIA, Altera, and Marvell. Most of these companies work with multiple foundries so they have even more exposure to differing process technologies than an IDM. So definitely we are fabless girls living in a fabless world!

My MBA is in Business Law to which I highly recommend. With the increasing legal issues the semiconductor industry is experiencing you will never be bored!

D.A.N.

 

[Daniel Payne]

I agree with D.A.N. because IDMs have always been very integrated between design and manufacturing, even at Intel in 1982 the fab guys asked design if we could live with Metal 1 routed only in one direction, while Metal 2 was routed in the orthogonal direction in order to improve yield. Design vetoed the idea in order to get maximum layout efficiency and achieve the smallest die size.

Design, Manufacturing, Test and Marketing held weekly meetings called GYAT (Get Your Act Together) - so communication was constant between the groups.

 

[Paul McLellan]

I still think that the biggest issue is what the cost per function is going to be going forward. We have never before had an industry where that cost was not decreasing exponentially (this is what Moore's law is really all about) but it is unclear if that will still hold going forward. We are so used to electronics just getting cheaper and with more and more functionality at a rapid pace but we may be at the limit in cost/transistor.

 

[Daniel Payne]

A few years back Aart de Geus of Synopsys coined the phrase techonomics to define what the semiconductor industry has to offer in terms of both technical and economic viability, so I agree with your point Paul.

 

[shaz]

Hi D.A.N
Sorry for not explaining the way I wanted to. I totally agree with you. When I was at Fabless we had a group dedicated to interfacing with TSMC and that group was THE interface to the foundry and was so well integrated on both sides. Now, I work at the IDM and it feels exactly the same. In a way, fabless created the very interface (that existed with in the IDM) with an added advantage of being able to interface with other foundries. This better integration is a melting IDMs further.

It is very interesting, one of the CEOs mentioned to me that one of the hot areas that I could pursue within MBA project is: Patent litigation: the impact of so called ‘Non Practicing Entities’ (NPEs or also called patent trolls) on the Semiconductor industry.
Intellectual ventures come into my mind for some reason. I am aware that this is a very interesting area at the moment, however, it might disappear in few years (not sure though). Michael also wrote about paying attention to BUSINESS LAW. Above all, the fact remains that my interest lies more in the strategic issues affecting the industry than the legal side.
Thanks for your valuable input. I will definetly be writing more and more on the semiwiki as I develop my project. At the moment, I am trying to nail down a problem to work on that would not only be valuable to the industry but also give me an opportunity to move into a non technical role which I am finding very hard to make a switch.

Thanks,
Shaz

 

[shaz]

Paul, this is exactly what I am trying to nail down. The technomics does not allow the cost of the chip increase. The cost is a function of the process node development, EUV (going forward), the wafer size and even the design prodcutivity gap. How will the players devise startegies to be profitable and not let the cost rise. More collaboration, or more consolidation (economies of scale), or slow down Moore. Strategy in retrospect is always easy to describe but extremely diffciult to create for the future and that is the reason, it is on the CEOs minds. The industry clearly is in transition.

 

[LinkedIn]

Sadayuki Yoshitomi • About the sitlation of the Japanese Semiconductor Industory, the Daniel's WEB article hits point. Japanese Semicondudtor Industories are no longer main players in the FD-SOI, FinFET industory. No further investment on such new developments will be made. My personal perspective is there are two keywords in the indostory the coming future; (COMMODITY) standing for foundry services (target process ; deep submicron CMOS, SOI, FinFET) by existing big players, (COSTOMIZATION) standing for IDM services (target process ; not advanced CMOS process) by others for high-end products.

 

[LinkedIn]

Ken Huening • With the industry maturing there are some predictable directions technology is taking. These are mentioned in the ITRS report (thanks for sharing this). Then there are some unpredictable directions looming. The semi market is not one market any longer. The report focuses on a lot of Flash, MPU and ASIC technologies where smaller is key. There are a number of other market segments less reliant on transistor size and speed. Clearly the size/speed segments are becoming a "red ocean" of more difficult profitability. Subcontracted wafer fabs, fab consortiums are a natural competition to captive fabs in these areas and they are growing. Global Foundaries grew 36% last year. Cost is only one competitive strategy. Michael Porter, in his Competitive Strategies text states that there are three key strategies.....cost, differentiation and focus. Cost is a hard one to win by in todays market. Many profitable companies are focusing on customers and product differentiation. The customer focus strategy is becoming more difficult. Apple and Samsung have realized that they dont want to pay their suppliers 70% margins on the parts they buy. They find capable engineers, because of industry downsizing, readily available to design, test, and roadtest their own captive designs. With outsourced fabs they can develop their own products and retain the IP. Differentiation will come from unique algorithms and niche processes. The demand for smart power, high power, high voltage, high temperature circuits still provides opportunities for companies to develop unique processes, circuits and test methods that are unique and command higher margins. Fabless companies will be hard pressed to invade these territories and they will be used in high volumes once the technologies mature. GaN and SiC offer some aspect of this for temperature, voltage, high frequency and lower resistances. Other companies will develop strong IP for specialized circuit applications, like switched mode DC/DC conversion done with low power and high efficiency. The world is run on batteries and making them last longer till there is a better alternative is a large market. Coming to a point though the market is consolidating and most of the former mainstream processes are being offered by subcontracted fabs. Companies have to look to what will differentiate their products to maintain margins.

 

[LinkedIn]

Daniel Tomaszewski • It's so expensive to develop and maintain an economically-effective process for fabrication of circuits based on very small-size devices, that only the biggest players can afford. On the other hand will the biggest players customize the process taking into account different needs ? The Moore law has lost its primary meaning as description of a trend, but has become something to force chipmakers to make more and more efficient circuits, and finally to narrow a space for their activity.

 

[LinkedIn]

Daniel Tomaszewski • I would like to modify slightly my last comment.
It's so expensive to develop and maintain an economically-effective process for fabrication of circuits based on very small-size devices, that only the biggest players can afford. On the other hand will the biggest players customize the process taking into account different needs ? The Moore law has lost its primary meaning as description of a trend, but has become a rule which forces chipmakers to scale technology more and more aggresively almost with no respect to costs. Hence, who can not withstand this pressure, fails. Only the biggest survive. Thus manufacturing becomes concentrated and e.g. Europe becomes fabless.

 

[buzunca]

Hi shaz and everybody,
I do my PhD in Strategy and I focus on semiconductor manufacturing industry.I share some academic articles below. Maybe they will help for your theory part and contribute to the discussion here.

Regarding how IDMs survived in the face of specialization and which types of innovations they focus on:
Rahul Kapoor (2012), Persistence of Integration in the Face of Specialization: How Firms Navigated the Winds of Disintegration and Shaped the Architecture of the Semiconductor Industry, Organization Science , (forthcoming).

Regarding Dominant technology adoption (from CMOS to Nanotechnology now):
Jiang, L., Tan, J. and Thursby, M. (2011), Incumbent firm invention in emerging fields: evidence from the semiconductor industry. Strat. Mgmt. J., 32: 55–75. doi: 10.1002/smj.866

Regarding the role of innovation challenges in a lithography equipment ecosystem:
Ron Adner, Rahul Kapoor (2010), Value Creation in Innovation Ecosystems: How the Structure of Technological Interdependence Affects Firm Performance in New Technology Generations, Strategic Management Journal, 31 (3), 306 - 333.

 

[shaz]

Ken Huening • With the industry maturing there are some predictable directions technology is taking. These are mentioned in the ITRS report (thanks for sharing this). Then there are some unpredictable directions looming. The semi market is not one market any longer. The report focuses on a lot of Flash, MPU and ASIC technologies where smaller is key. There are a number of other market segments less reliant on transistor size and speed. Clearly the size/speed segments are becoming a "red ocean" of more difficult profitability. Subcontracted wafer fabs, fab consortiums are a natural competition to captive fabs in these areas and they are growing. Global Foundaries grew 36% last year. Cost is only one competitive strategy. Michael Porter, in his Competitive Strategies text states that there are three key strategies.....cost, differentiation and focus. Cost is a hard one to win by in todays market. Many profitable companies are focusing on customers and product differentiation. The customer focus strategy is becoming more difficult. Apple and Samsung have realized that they dont want to pay their suppliers 70% margins on the parts they buy. They find capable engineers, because of industry downsizing, readily available to design, test, and roadtest their own captive designs. With outsourced fabs they can develop their own products and retain the IP. Differentiation will come from unique algorithms and niche processes. The demand for smart power, high power, high voltage, high temperature circuits still provides opportunities for companies to develop unique processes, circuits and test methods that are unique and command higher margins. Fabless companies will be hard pressed to invade these territories and they will be used in high volumes once the technologies mature. GaN and SiC offer some aspect of this for temperature, voltage, high frequency and lower resistances. Other companies will develop strong IP for specialized circuit applications, like switched mode DC/DC conversion done with low power and high efficiency. The world is run on batteries and making them last longer till there is a better alternative is a large market. Coming to a point though the market is consolidating and most of the former mainstream processes are being offered by subcontracted fabs. Companies have to look to what will differentiate their products to maintain margins.

Ken,
Thanks for sharing your thoughts. From your post it is clear that the way forward for players is: differentiation and focus (it may even be true for the pure foundry). The current incumbents have to adopt more 'blue ocean' strategies. The value chain needs analysis for capturing value. More than Moore!!

 

[shaz]

Daniel Tomaszewski • I would like to modify slightly my last comment.
It's so expensive to develop and maintain an economically-effective process for fabrication of circuits based on very small-size devices, that only the biggest players can afford. On the other hand will the biggest players customize the process taking into account different needs ? The Moore law has lost its primary meaning as description of a trend, but has become a rule which forces chipmakers to scale technology more and more aggresively almost with no respect to costs. Hence, who can not withstand this pressure, fails. Only the biggest survive. Thus manufacturing becomes concentrated and e.g. Europe becomes fabless.

Yes Daniel, the consumer and the system developer is hooked on seeing 29% reduction in chip costs per node and they are the winners in the trio. The chip makers are leaving money on the table which the system developer and the consumer is cashing on. The chip designers can take some value back, that value can be from the bundled software that goes with almost every chip.
Thanks for the input.

 

[shaz]

Hi shaz and everybody,
I do my PhD in Strategy and I focus on semiconductor manufacturing industry.I share some academic articles below. Maybe they will help for your theory part and contribute to the discussion here.

Regarding how IDMs survived in the face of specialization and which types of innovations they focus on:
Rahul Kapoor (2012), Persistence of Integration in the Face of Specialization: How Firms Navigated the Winds of Disintegration and Shaped the Architecture of the Semiconductor Industry, Organization Science , (forthcoming).

Regarding Dominant technology adoption (from CMOS to Nanotechnology now):
Jiang, L., Tan, J. and Thursby, M. (2011), Incumbent firm invention in emerging fields: evidence from the semiconductor industry. Strat. Mgmt. J., 32: 55–75. doi: 10.1002/smj.866

Regarding the role of innovation challenges in a lithography equipment ecosystem:
Ron Adner, Rahul Kapoor (2010), Value Creation in Innovation Ecosystems: How the Structure of Technological Interdependence Affects Firm Performance in New Technology Generations, Strategic Management Journal, 31 (3), 306 - 333.

I can not thank you enough for the useful articles that you have pointed. My project will be in a slight different direction than yours (reading your other post). Anyways, any thing related to strategy in semiconductor please feel free to share in my personal email or here. It is hard to find material on this topic.