Archive for 2011

Could TSMC be your next chip design cloud owner?

Tuesday, October 25th, 2011

In a 2009 technical report, ”The Clouds: A Berkeley view of cloud computing”, the authors cite “Cloud Computing is likely to have the same impact on software that foundries have had on the hardware industry”. The underlying logic being the high cost of semiconductor fabrication line leading to the rise of semiconductor fabs and these in turn “enabling” fabless semiconductor design companies whose value is in innovative chip design.

A week back, I moderated a panel discussion on cloud computing in the IC design world, especially on accelerating time to market. While issues on security – a challenge which is stacked right on top of the “barriers to entry”, were defended and discussed quite animatedly, Surprisingly, it was the “cloud ownership” aspect which evoked only some tepid responses.

Now from where I stand, I see cloud ownership as a vital component of chip design security in the cloud. After all, if I were to place my company’s most precious assets –i.e. my chip design database – on a cloud, I will definitely like to know as to who owns the cloud. And this is on top of my regular apprehensions about my data security, back up and related aspects.

Let me clarify - I am not talking here about the infra structure provider e.g. Amazon and the likes. Rather it is the cloud framework/database owner. The framework here includes components of the existing physical eco-system integrated together – design database, EDA tools, user interface etc. – without which cloud computing will just service individual IC design tasks i.e. storage and processing power requirements; something which on its own is not exactly fully leveraging this powerful biz paradigm shift aka cloud computing.

So again – who will own the chip design cloud? Will it be the foundries (also cited as “natural design aggregators”), the EDA vendors, the fabless design companies or yet another entity? The reply gleaned from most of the stormy discussions elsewhere in the nimbus zone gravitates towards foundry.

Which brings me back to the where I started this post – Riding on cloud computing, foundries may turn out to influence the hardware industry in more ways than one. And who else is better equipped to lead the way here than TSMC??

So what’s the deal with the Google –Motorola Mobility deal, eh?

Tuesday, August 16th, 2011

The cyberspace is abuzz with news about Google acquiring Motorola Mobility for a whopping $12.5 billion. Speculations on Google’s motive behind the deal are mainly skewed towards 2 issues – access to patents and the other whether Google has plans to set up another end-to-end mobile empire akin to Apple. Add to that the buzzing concern of a high potential for conflict - mainly whether there still will be a “level playing field” amongst different Android handset vendors.

My two cents….

1. Access to patents: Motorola deal gives Google access to more than 17000 patents. This helps Google lend a legal hand to the embroiled Android handset vendors like HTC and Samsung as well as prepare itself against potential infringement law suits.

Flip side: If that was the main point, would Google not have been better off with just buying the patents like it did from IBM?

2. Potential conflict of interest – open OS partner or a handset competitor?? Now that can indeed be a worrying factor amongst the Android handset vendors - in spite of the prompt support statements from Samsung, HTC etc. Will these companies who had flocked towards Android to compete against Apple now gravitate towards Microsoft’s Windows Mobile? Google has stated that the handset biz will be kept as a separate independent biz and Android platform as an open one as before but the company will have a tough time treading this slippery slope in order to retain the Android handset vendors’ support.

Having said that, we have seen biz areas where the line between partner and competitor has blurred. Pure play foundries, IP vendors and IDMs is one such example. Market conditions have led to consolidation, fab lite etc. IDMs get the core process wafers done from foundries and keep some special process add-ons in-house to retain their specific niche. IP vendors work along with foundries in spite of foundries touting their own IP portfolio as well as specific design services.
Individually it has got very difficult to compete, with combining resources, there always lurk the spectre of “loss of level playing field”.

3. An Apple like end-to-end empire: Lucrative but an extremely difficult path ahead for the search engine giant in the hardware world… a tough act to follow!

But apart from these, the news throws up another nugget too:

The central point of computing is moving away from the desk towards mobile - and search engines do follow the computing devices. A tight integration between hardware and OS will make it easier to get the desired utilities and apps to the consumer – providing the coveted “unique user experience’.

The deal goes beyond handsets. Motorola Mobility is also into set top boxes – just to name another one. This will bring Google back into the home automation market. Gigaom’s Stacey Higginbotham & Katie Fehrenbacher has written a very good article on this; do read it. Getting your ads, location optimized and perhaps with dynamic relevance does require a tighter integration between the hardware and software.

And it is for this reason alone, my opinion is that it will very much be in Google’s interest to keep Motorola Mobility humming away as a separate unit within Google – especially as far as the level playing field of Android handset entities are concerned and leverage this hardware vendor acquisition to bolster its search ads revenues by making it’s ads more pervasive and relevant.
The patents are the special icing on the cake!!

MStar & Mediatek vying for Nokia’s 2G biz??

Tuesday, July 26th, 2011

What happens when two local companies fight for the same pie? And when that pie has a component from a “once market leader, now floundering handset maker”? And when one of the pie contestants is also gnawing the market share of this “once market leader, now floundering handset maker”?

Well, speculations abound!

MStar and Mediatek,the two Taiwanese chip makers for mobile handsets are vying for Nokia’s 2G phones’ biz - a volume estimated to be around 285 million in 2013.

Mediatek made its prolific rise in handset chip biz by supplying chips to the shanzai/white box handset makers in China. By doing so, it eroded Nokia’s market share in that segment (of course, the other end of the spectrum was eroded by Apple, Samsung and the likes!)

MStar steered clear from this segment and has supplied customized chips to Samsung and LG for higher end mobile phones. Perhaps, these will steer Nokia’s biz towards them…

Lately Mediatek has been shedding it’s “shanzai market” tag and moving up the value chain – Spreadtrum may have a lot to do behind this move!

Interesting times ahead for these sparring compatriots!

Spreadtrum acquires Telegent - Deja-vu??

Friday, July 22nd, 2011

Spreadtrum Communications, the fabless developer of baseband and RF chips recently announced its acquisition of Telegent Systems, a developer of software and silicon for the reception of live broadcast television signals.

While trolling the net, I saw this article that gives a quirky feeling of déjà vu. The article’s contents basically go on these lines….

In 2007, the US Wi-Fi provider and GPS manufacturer, SiRF bought Centrality, a company with navigation & multimedia experience. Later, SiRF itself got acquired by CSR. In hindsight, industry analysts viewed the Centrality purchase as a bad move.

Now the money/invested parties part …Centrality’s major investors included Walden International and it had a NEA (New Enterprises Association) principle on its BoD. An NEA principle was also on the SiRF’s board. Baseline appeared to be - sell the company to a public company and for enough cash that would return the VC capital and perhaps a large profit at the same time for the investors.

Move on to today…

Spreadtrum buys Telegent. Telegent shot into fame (and profits) with its analog broadcast TV. Since then, it moved to mixed signal and then to digital – a realm with intense price competition and very low margins.

Telegent’s investors include New Enterprise Associates and Walden International. Telegent’s CFO was once SiRF’s CFO. Spreadtrum has an NEA principle on Board….. you get the drift?

Will Spreadtrum do a SiRF???

Missing component in the local semiconductor eco-system

Friday, July 15th, 2011

An MoU was signed between Silterra and Might Meteor last week. It was basically about kicking off a training program under a wider umbrella of human capital enhancement. Being a part of the training program – and the fact that the MoU ceremony was held during the time I was there (!) – I got the opportunity to meet the top management involved from the two organizations as well as got a special guided tour of Silterra’s fab facilities.

For those who do not know - MIGHT-METEOR was established in February 2002 as a joint-venture company between Malaysian Industry-Government Group for High Technology (MIGHT) and Multimedia Technology Enhancement Operations (METEOR) – a consortium of 11 public universities in Malaysia. It is directly under PM Ministry.

I have always appreciated the efforts (and the funds) put in by the various government entities to grow and catalyse the semiconductor and other electronics biz in the country – I get a big chunk of my training biz from there! But jokes aside, I have seen a consistent commendable effort being put in this direction over the last few years that I’ve been involved.

The country is more known for the packaging and testing parts of the semicon supply chain. Yes, there are chips and system design houses but their number and biz is smaller compared to the biz brought about the A&T and contract manufacturing facilities. Penang’s Bayan Lepas area is the industrial hub for several big names – Intel, Altera, Agilent, Motorola, Huawei, Fairchild, Renesas and others. Kulim, in the neighbouring state of Kedah is another sprawling area with fabs, A&T as well as offices of contract manufacturing companies like Celestica etc. Capital city KL has a few design houses as well as MIMOS.

During my various stints in the Bayan Lepas complex, I have attended a few IEEE talks too, courtesy the local IEEE chapter. These events mostly lasted no more than a couple of hours, after work, and did attract a decent size crowd – a good platform to share insights as well as network. The local Centre of Excellence in Electronics also holds regular technical talks. Different government-industry-universities consortia exist. Most of the components of a thriving semicon eco-system exist. What is missing though, especially for a country moving to high income, innovation economy is the buzz and the high energy that is usually associated with such an eco-system – making it vibrant. The one thing that intrigues me is that in spite of the ingredients, the region (Penang, Kulim) lacks any big annual semiconductor event. An event where industry/govt/academia leaders can share their insights, members of the eco-system can network, biz mapping be done etc. Neighbouring Singapore (where I am based) has regular shows like SEMICON, CommunicAsia, Electronics Fair, MIDAS Semicon Show (now SSIA semiconductor show) etc.

Would be great to have something on the above lines there….!

Strategy in today’s smart phone & tablet market

Saturday, June 25th, 2011

On a recent flight to Penang (a biz trip and not a vacation!), I was thumbing through a special edition of HBR – Essential reads for Global leaders” and came across this interesting and old HBR article “What is Strategy?” by Michel Porter. While this was published in HBR Nov-Dec 1996 issue, it reflected so much of today’s biz world, especially the hot competition amongst the smart phone and now the tablet vendors that I wanted to share it here.

First the article outlines the difference between Operational Excellence (OE) and Strategy – something that is so often confused in today’s biz arena. OE means performing similar activities better than rivals. Strategy is the creation of a unique and a valuable position, involving a different set of activities. It is creating a fit among its activities. The essence of strategy is not just of what and how to do but very much of what NOT to do.

Some statements from the article, which resonate so well in today’s market of smart phones and tablets –

“Gradually managers have let operational effectiveness supplant strategy resulting into zero-sum competition, static or declining prices ad cost pressures”

Take the example of Mediatek. By providing a cheap reference design and a complete system, it became a game changer for the shanzhai market in China. However, this was more of an OE coupled with targeting the right market. With Spreadtrum making inroads in the same space at even more competitive prices, Mediatek is now losing ground to such rivals and is tweaking its strategy.

“Competitive convergence is one reason why OE is insufficient. As rivals imitate one another’s improvements in quality, cycle times, or supplier partnerships, strategies converge and competition becomes a series of races down identical paths that no one can win”

“Strategic fit among activities is fundamental. It is harder for a rival to match an array of interlocked activities than it is merely to match a process technology or replicate a set of product features” – Now is this not exactly what Apple is doing (user experience pervades all activities within the company. Everything Apple does just re-enforces this and it has a close end-to-end system to ensure this)? The others are following it with a “me-too” approach, which can provide an explosive growth for an interim but not last long.

On approaches to preserve growth and re-enforce strategy – “One approach is to look for extensions of the strategy that leverage the existing activity system”

I look at RIM and its foray into tablets with Playbook here. RIM’s strength with its Blackberry phones has been enterprise. Analysts attribute RIM’s decreasing influence to the growing trend of CIO’s decisions on smart devices to be used by the enterprise employees as being increasingly based on what the employee uses as a consumer – thus skewing it away from the enterprise world which had always been RIM’s strength and essential to its strategy.

But then look at it from another perspective. Tablets are not just being used as a general consumer product – for the consumer to be connected. It has a tremendous potential (and is already being used thus) for enterprise productivity. RIM needs to see how to connect or rather leverage on this fit and communicate the strategy better to the customers who value this enterprise productivity.

Tablets and other MIDs: Commoditization of hardware??

Tuesday, June 7th, 2011

A Tablet market report from Goldman Sachs states “The OS platform wars could drive greater hardware commoditization over time. We believe that over time the more open platform vendors may have to impose standard hardware and user interface specs on handset and tablet OEMs to ensure that software developers have a uniform installed base. This move to standardization would narrow the ability for hardware manufacturers to differentiate their technology over time and could result in hardware commoditization like that found in the traditional PC market.”

With the gaining importance of software in the Mobile Internet Devices (MIDs), hardware’s role as a differentiating factor is indeed diminishing. And with that, so do the profit margins for the chip industry incumbents. So, how are the chipset players reacting to survive, if not thrive, in this evolving market?

Qualcomm shows a recent example - “Qualcomm will give web apps a boost”.
As a part of the company’s effort to enable a shift away from today’s fragmented set of native mobile environments, it is set to release shortly a set of applications programming interfaces geared to give Web-based applications deeper links into hardware. The company already supports Android, Blackberry, Windows Phone and WebOS mobile OSes among others. A move to Web-based applications would help it reduce the variety of platforms for which it needs to write software supporting its chips.

Web vs. native apps - as the mobile usage increases, both will grow with it and become valuable factors of product road maps. The question the product strategists need to ponder upon, however, is “what does my target audience need?” While the debate of web vs. native apps is not new, it does throw some interesting options in this backdrop of looming hardware commoditization.

One option is - The chipset vendors start conforming to the standard specs set by the open platform vendors. The hardware is strongly connected to the OS platform and with a proliferation of various mobile OS in the market, it is not an easy task supporting them all or even hedging on a few. Not enticing.

But what if a chipset vendor were to make inroads into web apps and get a deep link between web apps and its native hardware through some popular browsers? it can potentially get some interesting revenues by tapping the right web apps based on their target market – and remember that web apps is an open platform – no waiting, no approval. Its success is hinged on its adoption by the user community.

Having said that, the speed comparison (of compiled vs. interpreted code/web vs. native) will be there as well as cases, especially till the near future, where native wins over web but companies are working on those too (Qualcomm has been working for two years to optimize software so that browsers run as fast as possible on its chips). What has happened to desktop apps, can also happen to native mobile apps. Hmmm…. This may be one escape route from the commoditization problem.

India’s latest fab push

Monday, April 25th, 2011

Setting up fabs has been a point/initiative that keeps popping in and out in the Indian semicon landscape; and has been happening ever since India’s main fab (SCL) in Mohali had a major fire in late ‘80s!
(incidentally STMicro and IME/S’pore owe a lot to that as the SCL employees moved to these and other pastures following the fire!)

As mentioned in the article, a couple of years back the fab initiative again got some major push courtesy SemIndia as well as Hindustan Semiconductor Manufacturing Corp/Infineon combo but never took off.

While there are several proponents for building semicon fabs in India, personally I am not too gung-ho about the same J, mainly because

·        India does not have the apt environment and infrastructure needed for a fab (these fabs can help a lot on the local employment scene, though)

·        Fabs are too costly to be handled by small players. I do not see the big ones entering here in India w/o being incentivized by sizeable subsidies/tax breaks from the govt.  Intel was an interested player (though I think for backend fab) but the govt. and the tech bellwether could not come to anything mutually beneficial.
·        Even if the fab comes, the technology produced is another question. Are we talking about the low end technos (how will that develop “localized content and value-addition to the Indian made electronic products”?) or the high end ones (which IDM/foundry will opt for this?)
·        With fabs increasingly being seen as only a big players’ game (or of a consortium of big players) and IDMs going fab-lite, there is not much fiscal sense in pushing for a domestic fab in India. Even strategically (long term), it is a too high cost game for new players. However, should it be for some analog process (which generically trails behind the digital ones), it may make sense – will TI be interested (it already has a big R&D set-up in India)
 

 ·       I personally see hardware morphing into a diminishing role in the electronic products – software (embedded as well as application) is growing into a differentiating niche. India can leverage its strengths in embedded design, software development and large domestic market to attract the big global players further and climb up further in the value-chain. It has already been doing so (moving from routine tasks outsourced from the global HQs to local offices to chip development and now slowly on to system development)

 Domestic fabs can spawn the right eco-system to catalyse India into a big player but right now the odds are too much stacked against it – and am not sure if the projected gains are sufficient to warrant this path….

Oracle vs. HP and Intel’s Itanium – a semiconductor perspective

Sunday, April 24th, 2011

Recently a friend mentioned about the tremendous impact on Intel from Oracle’s March 22 announcement of its discontinuing development on Itanium.

While I had known about this announcement, I had not really given it much thought from the semiconductor perspective. So when I heard this comment, I started digging up…!

A bit on the background first: Using an architecture (based on explicit instruction level parallelism - in which the compiler decides which instructions to execute in parallel; as opposed to the superscalar architectures in which the processor manages the instructions dependencies at run time) that originated at HP and was later jointly developed along with Intel, Itanium is a family of 64-bit Intel microprocessors. The architecture was formerly called IA-64.

By 2009, the chip was almost entirely deployed on servers made by HP, which had over 95% of the Itanium server market share making HP-UX the main operating system for Itanium.

Ironically, the hardware competition to Intel’s Itanium comes from a chip from its own stables – the Xeon (Intel’s response to AMD’s x86-64 Opteron in 2004) that has cranked up its market share in the server pie over the years.

In comparison with its Xeon family of server processors, Itanium has never been a high-volume product for Intel. Majority of Intel’s server business is x86 and it will in the end suffer only a token loss of revenue as a result of Oracle’s announcement. Itanium has become such a niche product for Intel that the announcement had little effect on the chipmaker’s stock - Intel shares were up .02%, to $20.15 the morning after.

Here are some figures:

In the fourth quarter 2010, both Gartner and IDC saw x86 server revenues grow more than 20 percent. According to Intel (which cites figures by IDC), revenues of Itanium-based servers was $4 billion in 2010. By contrast, x86-based server sales was around $33.3 billion in 2010. With the x86 server market being eight times larger than that of Itanium, it does not make much sense for Intel to invest into Itanium’s development. On the other hand, it also cannot simply ignore the $4B Itanium market especially with the fact that the sockets for Itanium and Xeon are NOT interchangeable – as yet. Intel shared some information and re-iterated its commitment on the Itanium roadmap this month.

Btw, Intel recently announced 2 more server makers into its Itanium roster: Huawei Technologies and Inspur – that’s a 50 per cent growth rate in the Itanium OEM base.

IC Insight’s revised capex forecast

Wednesday, April 13th, 2011

IC Insights revised its capex forecast last week in which they indicate that semiconductor industry capital spending is projected to grow to $60.4 billion in 2011, up 17 percent from $51.8 billion in 2010,

I had done some number chewing from the earlier (Jan) report in a previous blog. So was curious to look at the changes. Looks like a couple of big ones in this revised one from the previous (Jan ‘11) forecasts - and they are for ST and Sandisk.

ST’s is 100% (from Jan’s 750M$ to April’s 1500M$) and Sandisk’s is 77.7% (from Jan’s 900M$ to April’s 1600M$). TSMC’s % change from previous is 23.8% increase.

Sandisk had already reiterated its 2011 capex budget of $1.4-$1.6 bn in Feb’11 and IC Insights’ new table reflects that (where did the earlier figure of 900M$ come from??). ST had also announced its capex plans ($1.1 -$1.5b) in as early as late Jan ‘11 (reflected in IC Insights’ latest table)

Japan quake and the semiconductor supply chain

Tuesday, April 12th, 2011

The infrastructure (transport, power etc.) disruption has indeed caused problems to the supply chain. The effect is not so immediate due to the inventory stock available with most customers. People here in Singapore while worried do not generally anticipate much problem in the semicon supply chain. Most of the companies here have enough inventory to last at least the next one to two months. In the meantime, they are exploring how to diversify their supply chain as well. Japan has however, is still experiencing aftershocks.

Having said that, I think that it will be the electronics supply chain (as compared to the semiconductor one) which will be more impacted. Japan is a big player in consumer electronics but not necessarily in semiconductors. (It accounted for 13.9$% of 2010’s global electronic equipment revenue and 16.5% of consumer electronics). Plus other than a TI fab, most others escaped damage.

But as Japan is the world’s largest supplier of silicon, raw wafer inventory will soon get depleted.
The infrastructure disruption is likely to cause short supply and higher ASPs (analysts foresee higher than expected global semicon revenue).  Affected parts are likely to include NAND flash, DRAM, microc, LCD panels and parts.

In fact on the electronics supply chain (a smooth semicon chain will not be of much help with a disrupted electronic one!), there was a recent interesting article in WSJ reporting a possible delay in iPod production because of the shuttering of a polymer (called PVDF and being used as a binder in batteries) manufacturing facility, owned by Kureha (it holds 70% of the PVDF market)., near the Japanese earthquake’s epicenter.  Reminds me of the old adage – for want of a nail, the battle was lost…

Analyzing some figures from “5 IC makers join $3B capex club”

Sunday, January 23rd, 2011

Analyzing some figures from “5 IC makers join $3B capex club”…..

Taking just the top 4 spenders and UMC -

Samsung made a quantum jump in total capital spending from $3.5B in ’09 to $9.6B in’10 – a 173% increase, It was the top spender in chip capex in ’10 (interestingly it’s ’10 capital spending was not too less than the combined cap spending of the next two entries in the list - Intel and TSMC). It’s ’11 forecast of $9.2B is a 4% dip from its ’10 spending – making it figure at the bottom of the 2011 major spenders ranked by forecasted spending change from the previous year. The absolute forecasted amount for ’11 is, though, still slightly higher than Intel’s $9B. However Samsung is known to outspend its target every year and analysts still predict a %-10% industry capex outlook.

The company is also cutting down on its DRAM spending. While overall semiconductor spending budget change is forecasted as 0% this year, the foundry spending is up (double of ’10) – still at record levels. Another interesting point to note is that in ’10 (a year in which % change in its capital spending from the previous year was a good 173%), Samsung’s capex/sales ratio (~38% with 9.6B capex and $25B sale, as estimated in mid ’10) was not too far off from its long term average of 32%.

Coming to Intel - Intel’s capex spending in ’10 was much less than Samsung for the same year ($5.2B vs. Samsung’s $9.6B) but is forecasted to be a massive one (73%) this year. However as pointed our earlier, the ’11 forecasted absolute amount will be slightly less than Samsung’s. This comes after a year which Intel called a record year and its best year. Major reasons cited for the industry bellwether’s increased capex in ’11 are - 22nm transition and expected increase in demand. Intel is approaching 22nm transition in ’11 and it also sees tremendous growth opportunities this year, especially with Atom based SoCs in smartphones, tablets, smart devices etc. as well as PC& server segments. Plus, it should not be forgotten that Intel is growing from 3 high volume leading edge fabs to 4.

Next the world’s top pure-play foundry – TSMC made major investment in’10 - a 120% ‘09/’10 % change amounting to $5.9B – slightly higher than Intel and much less than Samsung. However, the table indicates only a marginal forecasted ‘11/’10 % change – 7%, absolute sum of $6.3B. Note, though, that this is the highest absolute sum forecasted amongst all foundries.

Besides capacity, TSMC has always focused on being the technology leader amongst all the foundries. Recent news items point to TSMC’s high R&D spending in ’10 (the company’s R&D spending rose by 44% to reach $945M – moving it into the top 10 in R&D spending - and making it the first pure-play foundry to move into the top 10 semiconductor R&D spending club. TSMC hopes to catch a bigger share of the 28nm market this year.

Coming to GlobalFoundries – The y-o-y % changes are quantum (490% - ‘09/’10 and 96% - ‘11/’10) but then look at the starting point in ’09 - $466m which is quite low compared to TSMC and less than UMC’s too for that year – it had a lot of catching up to do. The foundry is in the process of expanding its current facilities (leading edge facility in Dresden/Germany), building the new manufacturing facility in NY and financing another production facility that will be located in Abu Dhabi.

The expansion focus in Dresden facility will be on adding new capacity to support additional growth opportunities for 45/40/28nm technologies as well as initial 22nm development. Target is to scale output to up to 80 thousand wafers/month over the next two years. The company has also been expanding its Fab 7 in Singapore to reach 50 thousand wafers/month across technologies ranging from 65nm to 40nm. GF’s 300mm output will be around 90 thousand wafers/month when all the new expanded facilities become fully operational.

Things are not looking too well for UMC – it started with a slightly higher capex than GF in ’09, spent half of GF’s (‘10/’09 % change) amount in capital in ’10 and is forecasted to keep the same capital spending in ’11. It could not catch up with TSMC especially on the technology front and now looks to be losing out to GF too. On a brighter note, UMC is reported to put CMOS –MEMS devices into volume production in 2011.

All these point to an intensification of competition on the foundry market as well as a rapid expansion of the contract manufacturers. The question is – will the industry see a foundry capacity glut in late 2011??