SemiWiki – Page 807 – The Open Forum for Semiconductor Professionals

RVN! 26 Banner revised (800 x 100 px) (600 x 100 px)

July 13, 2014

The iPhone6 will have TSMC 20nm, Absolutely!

The iPhone6 will have TSMC 20nm, Absolutely!
by Daniel Nenni on 07-13-2014 at 11:00 am
Categories: Foundries, TSMC

TSMC 20nm is one of the most talked about nodes on SemiWiki with lots of speculation and debate surrounding it. It’s interesting to look back at what we thought would happen to see if it actually did happen so let’s do that now. According to Google there are 411 articles referencing 20nm on SemiWiki, let’s visit a couple of the more interesting ones written by me (of course). This article is really for my beautiful wife and four children who rarely think I’m right so please don’t point out where I was wrong:

3D Transistors @ TSMC 20nm!
by Daniel Nenni
Published on 11-06-2011
Ever since the TSMC OIP Forum where Dr. Shang-Yi Chiang openly asked customers, “When do you want 3D Transistors (FinFETS)?” I have heard quite a few debates on the topic inside the top fabless semiconductor companies. The bottom line, in my expert opinion, is that TSMC will add FinFETS to the N20 (20nm) process node in parallel with planar transistors and here are the reasons why:

Dragon Boats and TSMC 20nm Update!
by Daniel Nenni
Published on 07-01-2012
Even more exciting, TSMC has 20nm up on the TSMC website now! Exciting for me at least! This is really cool stuff and it is right around the corner. I also like the new TSMC website and banner ads. It really does show a much more progressive communication style for a foundry…… Expect 20nm risk production to start in Q4 2013, two years to the quarter after 28nm.

Apple Will NOT Manufacture SoCs at Intel
by Daniel Nenni
Published on 12-09-2012
My bet is: moving forward Apple will use Samsung for 28nm (iPhone 5s) and TSMC for 20nm (iPhone 6). Intel certainly has a shot at 14nm and 10nm but never ever count out TSMC. If you want to bet a lunch on Apple manufacturing at Samsung or Intel for 20nm post it in the comment section. I will cover all lunch bets against TSMC.

TSMC 28nm and 20nm Update Q4 2012
By Daniel Nenni
Published on 12-16-2012
20nm will be a much more interesting node in regards to competition however. After learning the gate-first lesson, IBM is following TSMC with a gate-last HKMG implementation at 20nm. Unfortunately the added difficulty of 20nm double patterning and lithography challenges, which have yet to be solved at a production level, is causing delays. The fabless semiconductor ecosystem is working around the clock on this and I honestly expect a hockey stick 20nm production curve once this has been solved.

TSMC Apple Rumors Debunked!
by Daniel Nenni
Published on 01-11-2013
The first rumor is that the next Apple A7 processor (28nm) will be made by TSMC. That rumor is FALSE! As I previously blogged, the Apple iPhone to be released this year (iPhone 5s) will be Samsung 28nm. The iPhone to be released next year (iPhone 6) will be TSMC 20nm. A company the size of Apple cannot switch foundries on a moment’s notice. The volumes are too high and the technology issues are too complex. I have no doubt Apple discussed 28nm with TSMC but since no other foundries had 28nm available there was no way TSMC could handle the wafer demands of Apple and the rest of the fabless companies. Apple also gets preferred pricing so why would TSMC give up higher margin 28nm business AND alienate their customer base? Not going to happen….

Where will Apple Manufacture the next iPhone Brain?
by Daniel Nenni
Published on 07-17-2013
There still seems to be a lot of confusion here so let me set the record straight. In regards to the Apple Ax SoC, the Apple iPhone 5s will have Samsung 28nm Silicon. Samsung 28nm is still ramping but Samsung can make enough wafers and eat the yield issues no problem. The Apple iPhone 6 in 2014 will have TSMC 20nm as I reported previously.

Also read: Intel 14nm Delayed Yet Again?

Imec’s Process Secret Decoder Ring

Imec’s Process Secret Decoder Ring
by Paul McLellan on 07-12-2014 at 11:00 am
Categories: ESD Alliance, Semiconductor Services

To wrap up Semicon West, let’s go back to Monday and the imec presentations. In fact, An Steegen’s presentation titled The Semiconductor Roadmap. She covered a lot of ground, but some of her slides contain a wealth of information. Let’s look at the options for 10nm, 7nm and a little 5nm, what imec call N10, N7 and N5 (which makes sense since there isn’t anything 10nm on a 10nm process, it just represents a name that scales down from larger nodes where the dimension of the process really did correspond to the size of the transistors).

This graph above (click to enlarge) is a little complex but worth understanding. What it shows is the various tradeoffs we can make at each process (the green lines) on metal2 pitch and contacted poly pitch. The horizontal and vertical lines show the limits of lithography: 193i LE3 (aka litho-etch-litho-etch-litho-etch, so triple patterning), SADP (self aligned dual patterning, also sometimes called sidewall image transfer), single exposure EUV (so single patterning), 193i SAQP (self aligned quadruple patterning) and high numerical aperture EUV single exposure. We would like to keep the metal pitch above as many of the the horizontal lines and the CPP pitch to the right of as many vertical lines as possible.

If you extend that approach to a lot more layers then you end up with the table above (which assumes no EUV). Note that on metal1 it is assumed at N10 you can have two-dimensional metal (metal can run horizontally and vertically) but that at N7 it is assumed to be one-dimensional (either horizontal which is a bit simpler but has an area impact since 7.5 track standard cells are not possible, or vertical which is more complex but has better area since 7.5 track standard cells are feasible, and some other stuff; click on the picture on the right for more details). There is a lot of information in this table. For example, at N10 the fins require self-aligned double patterning with a single patterned cut mask. At N7 it gets more expensive: self-aligned quadruple patterning with an LELE (double patterned) cut mask. You can see the mask counts at the bottom (just for the first few layers, more depending on how many layers of metal get put on top of that).

This table assumes EUV happens at N7 and is used in a hybrid process along with 193i immersion lithography too. EUV is used to avoid almost all the block and cut masks and so reduces the number of masks significantly.

One advantage imec has over most of us is that they are knowledgeable about many of the decisions made by the various manufacturers at each process generation. There is not only one answer. For example, at 16nm some people have apparently opted for 64nm metal pitch and others for 58nm, which you can also read off the table (the white box at the top of the dark blue bar). The above graph shows those choices. Actually only the white boxes on the tops of the bars are really relevant, they show the range of values that are possible at that node for the 3 parameters, the FinFET spacing, contacted poly spacing and metal pitch.

As I said, there is a huge amount of very detailed information contained in these graphs and tables, it is the magic decoder ring for the semiconductor process roadmap.

More articles by Paul McLellan…

July 11, 2014August 22, 2024

GlobalFoundries: Upcoming Events

GlobalFoundries: Upcoming Events
by Paul McLellan on 07-11-2014 at 9:01 am
Categories: Foundries, GlobalFoundries

GlobalFoundries is running a series of technical seminars in the US and Europe, following on from the successful versions run last month in Shanghai and Hsinchu. Among the topics that will be discussed are Foundry 2.0 and Collaborative Device Manufacturing for both Leading Edge and Mainstream technologies. The European events are in early October followed by the US events, as in the table below.

GlobalFoundries also has around 130 openings in Malta NY at Fab 8, and have announced that they will be hiring as many as 800 people this year in total. With the recent announcement that GlobalFoundries will be running an identical process to Samsung at 14nm, then this fab is going to be running a truly state-of-the-art process and I’m sure some high volume customers will be using both fabs for security of supply.

Also, coming up on Monday July 14th is not just National Day for France bit it is interview week in Malta. The GlobalFoundries job board, including a link if you want to attend interview week, is here. There are opening for pretty much all positions involved in running a state-of-the-art fab.

If you want to see a CNBC video about it from a couple of months ago then it is here. I don’t seem to be able to embed it. But here is a transcript anyway:you know malta is about 20 miles north of the state capital of albany and it’s not the first place that you would think of when you hear high-tech manufacturing but there is a 3 million square foot campus with a massive chip making facility here. and this year, global foundries plans to hire 600 to 800 new workers adding to its 2200 person work force. all at good salaries. here’s mike bruiseso. i can tell you that an average salary high to low end, is about 90,000 for a rollup for a salary. so what kind of workers are they looking for? all kinds. they need engineers with four-year degrees, they need technicians with associates degrees that are basically the mr. fix its of this manufacturing process. now, we can’t show you the real facility because it’s a clean. we’re not allowed in here. before anyone starts working here the head of technical training says they have to come to this lab to learn about the thousand step process it takes to make a single semiconductor and about the multimillion-dollar machines it takes to run that process. it allows them to learn how to use workmanship skills, hand tools, things like that, and then it also takes them through a trouble shooting component through the megatronics portion, learning how electronics, mechanical, pneumatics, all these things tail together to create an assembly line. now when global fountries opened the plant in 2012 most of the workers were from overseas but it’s been working with community colleges and local high schools to basically start classes that would help them get the workers they need with the skills they need. global foundries wouldn’t be here if it wasn’t for an aggressive tax package and aggressive sales pitch from the state of new york, though, and we’re going to have more on that story

More articles by Paul McLellan…

July 11, 2014December 27, 2023

CEVA + Riviera Waves = Complete WiFi and BlueTooth Solution

CEVA + Riviera Waves = Complete WiFi and BlueTooth Solution
by Eric Esteve on 07-11-2014 at 3:18 am
Categories: Ceva, IP

Thanks to the recent acquisition of Riviera Waves, CEVA is expanding from Digital only DSP IP, to supporting complete solution (Digital + mixed-signal) for both WiFi and Bluetooth functions. Such a move is already a very good market positioning, as customers prefer buying to a single source the digital and the mixed-signal part of the same function, WiFi or Bluetooth. We think this acquisition will strongly help CEVA to attack new market segments where connectivity through a wireless interface is key: Smart Home, Internet of Things, Wearable or Medical applications. The acquisition of Riviera Waves by CEVA for $19M (including certain performance and other milestones as well as a two years retention plan for current RW employees) is, at first, a success for a 29 employees company founded in 2010, reaching $3.6M revenue and break-even in 2013. Let’s take a look at Riviera Waves IP port-folio, and how it completes the offer from CEVA.

The WiFi demonstration board on the above picture helps visualizing the complementarity: you can see on the middle of the larger board the Riviera Waves MAC, developed on the CEVA Teak-Lite 4 DSP, connected to the (Riviera Waves) Hard Wired Modem. The Modem is interfacing with the (Riviera Waves) RF Module on the daughter board. This solution addresses the WiFi, from 802.11n up to 802.11ac 2×2. If you add a CEVA XC4210 modem to the above, the complete IP will support WiFi 802.11ac 4×4.

Bluetooth Low Energy (Smart Ready) BaseBand IP

Bluetooth Low Energy (Smart Ready) Modem and RF IP

The two IP (Baseband and Modem/RF) supporting Bluetooth 4.0 or Bluetooth Low Energy (BLE) from Riviera Waves could be used along with CEVA Bluetooth software stack. CEVA will be able to propose a completely integrated solution: Software Stack + Controller (Baseband and Modem) + RF, the customer selecting the most suitable option for the baseband, between RTL of Teak-Lite 4 DSP based.

If we move up to the system level, we can see that the offer from CEVA has been strongly enhanced, as the IP vendor can build a complete offer, Controller + PHY or Digital plus mixed-signal for WiFi and Bluetooth. Let’s take a look at the business impact by market segment.

Mobile (Smartphone, Media tablet, Ultrabook)
As of today, Bluetooth and WiFi functions were supported by specific IC. If we look at a chip vendor like Qualcomm, selling the main Application Processor, WiFi IC and BT IC, the trend is clearly to integrate the last two functions into Snapdragon. This means that the followers will also tend to integrate WiFi and BT in the Application Processor. But, unlike Qualcomm who bought Atheros to support these wireless interface, these chip vendors will have to externally acquire WiFi and BT as an IP. CEVA will be well positioned to serve this mobile market, as the company has a long presence through the 2G, 3G and LTE baseband DSP IP offer!

Smart Home
This market is still emerging, but we can guess that it is seen as a new opportunity to develop IC to address needs for various applications. If we think about any of these applications, it’s likely that internet protocol is over sized, leading to more complex, and more expansive solution than when using WiFi or Bluetooth. Such an emerging market could generate several dozens of opportunities per year and motivates start-up creation. We can foresee that low power consumption will be a strong requirement, pushing for WiFi or BT integration into a System-on-Chip. CEVA should be well positioned on this market. Just a remark: because CEVA’s business model is based on up-front license plus royalties, if it takes longer to reach very high production volume on this market, the license part of the IP vendor revenue should be already strong, if the company succeeds in penetrating this market.

Wearable, Medical
The same remarks made about Smart Home applies for the wearable and medical applications: WiFi or BT are well suited.

What could be the upside revenue generated by Riviera Waves? We have to take into account that RW products will benefit from CEVA brand name recognition, necessarily higher than a four years old company, as well as from a long time established, world wide sales channel. If we evaluate the SoC design starts for Smart Home, Wearable, Medical and IoT to be above 100 in 2018 (conservative), then a 20% penetration would generate 20 WiFi or BT IP design-win for CEVA. Adding Mobile segment and the related SoC design starts, would generate another dozen IP design win. It is not over-optimistic to foresee an upfront license upside potential in the $15M to $20M range in 2018… CEVA acquisition of Riviera Waves look like a good move!

Eric Esteve from IPNEST –

EDA Mergers & Acquisitions Wiki Now Includes IP

EDA Mergers & Acquisitions Wiki Now Includes IP
by Ian Getreu on 07-10-2014 at 4:00 pm
Categories: EDA, IP

When I started collecting my list of EDA mergers and acquisitions about 30 years ago, my objective was simply to determine the number of logos each EDA company owns. For that reason, I collected all the fish-eat-fish-eat-fish…. mergers and acquisitions, going way back to the very early days of EDA (even before it was called EDA). I represented these using indents. If an EDA company did not merge with or buy another company, they were not on my list. I tried to be very diligent over the years in keeping everything up-to-date.

For those interested in the present tallies of the big 3, I calculate that Cadence owns 109 logos, Mentor owns 113 logos and Synopsys owns 153 logos. No conclusions to be gleaned from this – just interesting numbers.

I did this as an individual – there is no official sanction for my list from any company or any organization. I can easily have missed mergers. Errors and omissions are mine, and mine alone.

In January, 2011, Daniel Payne suggested I put the list up on Semiwiki. Daniel posted it on 16 January 2011 (https://www.legacy.semiwiki.com/forum/showwiki.php?title=Semi+Wiki:EDA+Mergers+and+Acquisitions+Wiki) and it has had 38,388 views as of 9 July 2014 – making it, to my total surprise, the most-viewed post on Semiwiki. Daniel and I keep it up to date and we really appreciate your interest in it.

Since I started the list, EDA has changed (not surprisingly). The list has survived most of the changes. However, one change that has not been completely reflected in the list is the incorporation of IP into EDA. If established EDA companies (like Cadence, Mentor and Synopsys etc.) bought IP companies, these have been included (to the best of my knowledge). However, I have usually ignored IP mergers – especially when an IP company merged with or bought another IP company; such as the recent CEVA purchase of RivieraWaves.

Daniel and I have decided to now include IP companies in the list, so we have added CEVA-RivieraWaves. However, because of my past decision to ignore IP companies, the list may now be incomplete in this area.

We are therefore asking people who know about IP-IP mergers or IP acquisitions in the past to add them to our list if they are missing. It would be nice to make the list complete in this area.

Thank you – and keep enjoying the list.

P.S. I love the image I found – each fish carries its own baggage 🙂

July 10, 2014

Intel 14nm Delayed Yet Again?

Intel 14nm Delayed Yet Again?
by Daniel Nenni on 07-10-2014 at 12:00 pm
Categories: Foundries, Intel Foundry

This week I’m at SEMICON West with 27,000 of my closest friends. Good information, good networking, and some great rumors this year. Yesterday I heard a juicy rumor in the halls that Intel is still having 14nm yield problems. Remember, we heard a similar rumor last year and it turned out to be true. I Googled around this morning and found this:

Intel’s Broadwell Chips for Most Macs Not Shipping Until Early to Mid 2015

Intel’s line of 14-nanometer Broadwell chips, which are expected to be included in future versions of the MacBook Air, Retina MacBook Pro, and iMac, have been further delayed, reports Chinese site VR Zone.

According to the translated version of the article Intel has begun production on its low power Core M processors but production on the Broadwell chip series will not begin until much later in 2014. If that is in fact true then Mac shipments will be pushed out until 2H 2015, meaning it will not be a Merry Mac Christmas for some.

The Intel 14nm yield saga started at about this time last year:

Intel 14nm Delayed?
by Daniel Nenni
Published on 07-31-2013 08:45 PM
One of the more interesting pieces of information I overheard at SEMICON West earlier this month was that Intel 14nm was delayed. This rumor came from the semiconductor equipment manufacturers and they would know. What I was told is that the Intel 14nm process has not left the OR development facility to be replicated in the OR and AZ fabs….

No Mention of 14nm at the 2013 Intel Developer Forum?
byDaniel Nenni
Published on 08-19-2013 03:00 PM

Intel Really is Delaying 14nm Move-in. 450mm is Slipping Too. EUV, who knows?
by Paul McLellan
Published on 08-24-2013 01:23 PM

Intel Comes Clean on 14nm Yield!
by Daniel Nenni
Published on 12-04-2013 06:00 AM

Intel 14nm Delayed Again?
by Daniel Nenni
Published on 02-12-2014 07:00 AM

Intel 14nm Delayed Yet Again?
by Daniel Nenni
Published on 7-10-2014 10:00am

And yes, as petty as I am, I was rooting for 14nm problems back then as a result of Intel’s initial arrogance in entering the foundry business. Remember, EEETimes in May 2012 Intel started with “Fabless model is Collapsing” and “Fabless companies will not be able to follow where Intel is going.” Intel added that Qualcomm would not be able to use the TSMC 20nm process as an example. Well, Qualcomm is already shipping TSMC 20nm chips and has taped out FinFET versions of their mobile products.

Thankfully Intel has since softened its position against the fabless semiconductor ecosystem and I really appreciate that. Some people, like myself, have put their lives into this industry and we do not want to hear from anyone that we have failed when clearly we have not. Collaboration is key to success and Intel has a lot to offer us, absolutely. Intel’s next public address is on 07/15/14 at 2:00 PM PT Q2 2014 Intel Corporation Earnings Conference. Let’s hope there is transparency here so we can figure out what is really going on and learn from it for the greater good.

More articles by Daniel Nenni

July 10, 2014

Bob Metcalfe Keynote at #semiconwest

Bob Metcalfe Keynote at #semiconwest
by Paul McLellan on 07-10-2014 at 9:00 am
Categories: ESD Alliance, Semiconductor Services

My PhD is in distributed file systems so one of the key networking papers was Metcalfe and Boggs, 1976. It was titled Ethernet: Distributed Packet Switching for Local Computer Networks. This was the paper that introduced the world to Ethernet and to Bob Metcalfe. He was on stage yesterday here at Semicon West to give the afternoon keynote. Dressed in a suit and tie…and bright red running shoes. The talk was titled Innovation with Startups Out of Research Universities but it was really a little retrospective of Bob’s life and his ideas about how to foster and manage innovation.

In the early part of he worked on Arpanet and Project Mac at Harvard where he (eventually, he failed the first time) earned his PhD. He then worked at Xerox PARC during its heyday when a good fraction of the top computer scientists in the world worked there. PARC invented the Alto, which was a personal computer with a GUI and a mouse (although the mouse was actually invented by Doug Englebart up the road at SRI). Bob was given the task of linking all these computers together and worked with David Boggs, a graduate student. They took some of the ideas of the Aloha project (which linked the campuses of the university of Hawaii spread around different islands) and adapted them for wired connectivity and this was Ethernet. Bob pointed out that it would have been nice to use radio even then (WiFi 20 years earlier) but radios were too big, too slow and too expensive.

Bob left PARC and founded 3Com to manufacture networking equipment for the growing demand for both local area and wide area networks. Bob then became a venture capitalist and most recently became professor of innovation at the university of Texas in Austin.

Bob reckons that successful technology startups require three things:
[LIST=1]

People with the idea and the drive to implement the idea. Most ideas will turn out to be failures, even more so if someone other than the person who came up with it is not also the implementer. I like to say that the best way to transfer technology is in people’s heads. Research is almost all done at the top 100 universities in the US so the people with the idea will almost always turn out to be grad (or postdoc) students.

People with the skill to scale a company, preferably having done it before.

Money. VCs, angels, corporate investors. In fact Bob reckons that the best in an early stage startup is some money from the financial industry, some from the university where the inventors graduated, and some from a teaching customer who wants the product.

Bob also talked about the Bayh-Dole act. Prior to 1980, if your research was funded by the federal government then all the intellectual property rights had to be assigned to the government. This turned out to be a bad idea since the government is hopeless at actually capitalizing on it. The Bayh-Dole act changed the rules and the rights would instead be owned by the university employing the researchers. So all the universities set up technology licensing arms but they have been a mixed success since their goals are diffuse from benefiting humankind, making the most licensing revenue, making the product most likely to succeed (primarily by back-licensing it to the inventors). Bob reckons Bayh-Dole didn’t go far enough and it would have been much better to let the rights be retained by the professors and students who did the research.

Oh, and before you think that all these predictions are accurate since they come from an industry luminary, Bob has famously had to eat his words before. In 1995 he predicted that the internet would suffer a catastrophic failure in the following year. When it did not, at the 1999 conference on the worldwide web he took a sheet of paper with his prediction on, put it in a blender with some water, and literally “ate his own words.”

Earlier this year Bob did an AMA on Reddit All Your Ethernets Belong to Me. It is here.

More articles by Paul McLellan…

July 10, 2014

Are iPhones to Die For?

Are iPhones to Die For?
by Daniel Nenni on 07-10-2014 at 3:00 am
Categories: Mobile

First there were car phones and admittedly I waited in line to get one when they went mainstream. It was a Motorola something or other and it came with a curly antenna for the back window. From there I got a Nokia, a flip phone, a BlackBerry, and now iPhones. These life changing devices would not have happened without the fabless semiconductor ecosystem so congratulations to us all! Seriously, we changed the world, absolutely.

Unfortunately there certainly are downsides to new mobile technology. Distracted driving is one and we now have laws to help prevent that. In fact, I lobbied for them after I was run down while riding my bike and left for dead by a young lady who was arguing with her boyfriend on her cell phone. I was training for the Ironman Triathlon and as a result of the accident I can’t even do a Triathlon much less the famed Ironman. But I’m alive and that is more than some people can say.

The most disturbing trend I see now is people dying trying to get their phones back. A passerby asked a 15 year old girl what time it was. She took out her phone, he grabbed it and drove off (it’s called apple picking). She jumped on the car as it sped away to get her phone back and died when she fell off. This apple picker was caught and is being charged with murder. Another gentleman dropped his phone on the subway track and died trying to retrieve it. The most horrific one I read about is about a man who dropped his phone in a septic tank and was overcome by the fumes when he tried to retrieve it. Seriously, a septic tank? A toilet maybe, but a big tank full of nastiness? The find my iPhone feature is very useful but what if you find a bad person with it? My son told me a story about retrieving his fiancés iPhone after tracking it to the apple picker’s apartment. The picker sheepishly gave it back but that could have ended badly.

Are these devices really worth dying for? I highly doubt it is the cost of the devices that makes us risk our lives for them. I suspect it is the information they contain or, more likely, smartphones are now an important part of our hierarchy of needs. I added semiconductors to this one as a joke but look at the needs part of Self-actualization, Esteem, Love/Belonging, Safety, and Physiological. It is a bit frightening to see how big of a role technology now plays in this whole theory. Know what I mean?

I now tell my loved ones that replacing a smartphone is much easier than replacing them….. Maybe smart watches ARE a good idea?

Speaking of wearables, catch me at the CASPA Summer Symposium this weekend at the Intel Auditorium for: Enabling Technologies That Will Shape The Next Wearable Applications. My beautiful wife will be there too. I hope to see you!

Chip side of the Open Interconnect Consortium

Chip side of the Open Interconnect Consortium
by Don Dingee on 07-09-2014 at 9:00 pm
Categories: Foundries

Maybe it’s my competitive analysis gene, or too many years spent hanging out with consortium types, but I’m always both curious and skeptical when a new consortium arises – especially in a crowded field of interest. The dynamics of who aligns with a new initiative, and how they plan to go to market compared to other entities, prompts deeper exploration. Continue reading “Chip side of the Open Interconnect Consortium”

July 9, 2014

Mark Adams Keynote at #semiconwest

Mark Adams Keynote at #semiconwest
by Paul McLellan on 07-09-2014 at 12:58 pm
Categories: ESD Alliance, Semiconductor Services

The first surprise of the opening keynote for Semicon West was on the slides that were cycling on the screen as the room filled up. Somehow our book Fabless had managed to be in the rotation.

The opening keynote was by Mark Adams, the President of Micron. He was talking about upcoming big changes in the semiconductor environment, although given his background there was an especial interest in memory. Memory has been a weird market. From 2002-2012 nobody made any money. It was basically a commodity DRAM market and the companies in it, of which there were over 40 at the beginning of the era, overbuilt capacity leading to a price war. At the current node and with the worldwide downturn, the number of companies was down to 4 (Toshiba, SKHynix, Micron and Samsung) and they didn’t overbuild capacity. Indeed, a large part of the growth of the semiconductor industry in the last year or two has been simply the firming of prices in the DRAM market.

He sees several big challenges going forward, some of which are unique to memory but mostly not:

the end markets are getting more diverse. The main market a decade ago was PC, that has already been eclipsed by mobile, and whatever the internet of things turns out to be it will for sure be consist of hundreds of products
increased complexity of system level solutions, particularly 3D and novel packaging
decisions about foundry vs captive. Well, for everyone except the memory manufacturers and Intel (and, I suppose, Samsung) that decision has already been made
the fundamental semiconductor technology is at a crossroads, running into some fundamental limits requiring extensive innovation. In particular, for Micron, the upcoming end of planar in DRAM.
changing capital allocation models making capital harder to deploy and raising the importance of just how it is deployed (not to mention the cost of fabs means more capital is needed to build one)
economics don’t scale like they used to, a process node shrink does not generate as much return as before and so shrinking is less important than other innovation
finally, the customer landscape is evolving, moving (for memory) from just PC and networking towards embedded, server, mobile, client storage and more

A new more cooperative partnership model is required since no one company can do it on its own and since even memory, never mind SoC, is no longer a commodity.

It is obviously more exaggerated for memory, but all semiconductor business face the fragmentation of their market. They will need to partner with their customers closely to ensure they build the right products, but also they will inevitably be building a broader portfolio of products with relatively lower volumes for each one.

And developing the basic semiconductor process will also require partnering to do the development, to develop the equipment needed to manufacture, and to develop the new materials that will be required.

Success will depend on 5 factors:
[LIST=1]

Safety

Speed to market and cycle time, leading to faster learning

Quality. It has always been important but with increasing electronics in cars/medical etc it is more so than ever

Time to mature yield

And, of course, cost (BOM for some low end smartphones is now as low as $18, for example)

More articles by Paul McLellan…