Archive

Archive for the ‘Fab-Lite’ Category

Semiconductor supply chain dynamics: Future Horizons @ IEF2011


The last decade heralded a dramatic transformation in supply chain dynamics, driven by the complexity challenge of staying on the More Moore curve. On the demand side, the high cost of fabs persuaded almost all integrated device manufacturers (IDMs) to use foundries for their leading-edge wafer supply.

The ever-increasing process complexity and its negative impact on manufacturing yields forced the adoption of sophisticated foundry-specific design-for manufacturing (DFM) techniques, effectively committing new chip designs to a single foundry and process.

At the same time, the industry adopted a much more cautious lagging rather than leading demand approach to new capacity expansion, resulting in under-supply and shortages in leading-edge wafer fab capacity. To make matters worse, the traditional oxide-based planar transistor started to misbehave at the 130nm node, as manifested by low yields and higher than anticipated power dissipation, especially when the transistors were supposed to be off, with no increase in performance, heralding the introduction of new process techniques (e.g., high-k metal gates).

Even before these structural changes have been fully digested, supply chain dynamics have been further disrupted by the prospective transition to 450mm wafer processing, to extreme ultra violet (EUV) lithography, and from planar to vertical transistor design.

Transistor design
Since the start of the industry, adding more IC functionality while simultaneously decreasing power consumption and increasing switching speed—a technique fundamentally known as Moore’s Law—has been achieved by simply making the transistor structure smaller. This worked virtually faultlessly down to the 130nm node when quite unexpectedly things did not work as planned. Power went up, speed did not improve and process yields collapsed. Simple scaling no longer worked, and new IC design techniques were needed.

While every attempt was made to prolong the life of the classic planar transistor structure, out went the polysilicon/silicon dioxide gate; although this transition was far from plain sailing, in came high-k metal gates spanning 65nm-28nm nodes. Just as the high-k metal gate structure gained industry-wide consensus at 28nm, it too ran out of steam at the 22nm-16nm nodes, forcing the introduction of more complex vertical versus planar transistor design and making the IC design even more process-dependent (i.e., foundry-dependent). Dual foundry sourcing, already impractical for the majority of semiconductor firms, will only get worse as line widths continue to shrink. Read more…

On possible Samsung-SanDisk deal; AMD’s fab-lite path

September 6, 2008 Comments off

Last week, the global semiconductor industry has been hearing and reading about two big speculative stories:

a) A possible acquisition of SanDisk by Samsung, and
b) A possible chance of AMD taking the fab-lite route.

First on Samsung’s buyout (possible) of SanDisk! There have been rumors of a possibility of Samsung acquiring SanDisk. While it is still a possibility, it also leads to several interesting questions!

Should this deal happen, what will be the possible implications for the memory market? Will this also lead to a possible easing off on the pricing pressures on the memory supply chain? And well, what happens to the Toshiba-SanDisk alliance?

A couple of weeks back, iSuppli, had highlighted how Micron had managed to buck the weak NAND market conditions, and was closing the gap with Hynix in Q2, and that NAND recovery was likely only by H2-2009.

I managed to catch up again Nam Hyung Kim, Director & Chief Analyst, iSuppli Corp., and quizzed him on the possible acquisition of SanDisk by Samsung.

A caution: Remember, all of this is merely based on speculation!

On the possibility of Samsung’s takeover of SanDisk, he says: “Samsung at least said that they consider it. Thus, it is a possible deal. But who knows!”

Kim is more forthright on the implications for the memory market, should this deal happen, and I tend to agree with him.

Consolidation inevitable; no impact on prices
The chief analyst quips: “The NAND flash market is still premature and there are too many players in flash cards, USB Flash drives, SSD, etc. The industry consolidation should be inevitable in future.”

So, will this possible buyout at least ease some pricing pressures on memory supply chain? “I don’t expect this deal to impact the prices. Prices will depend on suppliers’ capacity plans. In the memory industry, the consolidation has never impacted the prices in a long run. (maybe, just a short-term impact). As you know, Micron acquired Lexar a few years ago, but no impact,” he adds.

Is there any possibility of SanDisk delaying its production ramps and investments at two of its fabs? And, what will happen should it do so?

Nam says: “SanDisk has already said that they would delay its investment and capacity plan given difficult market condition. This is a positive sign to the market as we expect slower supply growth than expected in future. However, in a long run, consolidation won’t impact the market up and down.”

Negative impact likely for Toshiba?
Lastly, what happens to the SanDisk-Toshiba alliance, should the Samsung buyout of SanDisk does happen?

Nam adds: “It is negative to Toshiba. The company [Toshiba] not only loses its technology partner, but also loses its investment partner. It should be burden for Toshiba to keep investing themselves to grow its business.”

Well, in SEMI’s Fab Forecast Report, there is mention of how Toshiba and SanDisk are among the big spenders in fabs, in Japan. Considering that Japanese semiconductor manufacturers are more cautious, it would be interesting to see how this deal, should it happen, affects the Toshiba-SanDisk alliance.

Now, AMD goes fab-lite?
While on fabs, this brings me to the other big story of last week — of AMD going the fab-lite route, possibly!

Magma’s Rajeev Madhavan had commented some time back that fab-lite is actually good for EDA. It means more design productivity. Leading firms such as TI, NVIDIA, Broadcom, etc., are Magma’s customers.

Late last year, Anil Gupta, MD, India Operations, ARM, had also commented on some other firms going fab-lite! Gupta pointed out Infineon, NXP, etc., had announced Fab-Lite strategies. Even Texas Instruments was moving to a Fab-Lite strategy. “IDMs are going to be the fabless units of today and tomorrow,” he added.

So much for those who’ve taken the fab-lite route, and industry endorsements.

On the fab-lite subject, iSuppli’s Kim will not speculate whether AMD would actually break up into into two entities: design and manufacturing, and also prefers to wait and watch.

How does fab-lite actually benefit? He comments: “Fab-lite has not been working well in the memory industry, which requires very tight control. It works, IF two companies (an IDM and a foundry) work very closely. For example, the industry leader, Samsung, produces all of the memory alone without any foundry relationship.”

Watch this space, folks!

Semicon to grow 12pc in 2008: Future Horizons

May 14, 2008 Comments off

If there is going to be a global economic recession, the chip industry (but not all companies) is in the best shape possible to weather the ensuing storm!

According to Malcom Penn, CEO, Future Horizons, we are dealing with a semiconductor industry in ‘deep trauma.’ He was delivering the company’s forecast at the recently held International Electronics Forum (IEF) 2008 in Dubai, predicting a 12 percent growth this year despite signs of a wobbling US economy.

Is there a need to get back to the industry basics? “Semiconductors are a peculiar business; the only sane strategy is to bet the company regularly,” once remarked Dr Gordon Moore.

Penn noted that the current industry status is somewhat confused and uncertain. Short-term issues are dominating the agenda.

Longer-term structural trends are unclear. The traditional IDMs are currently going through a mid-life ‘new business model’ identity crisis, and the start-ups are struggling to even reach critical mass! And all of this has been happening amidst intense economic uncertainty

“Now is the time for strong nerves and determination,” Penn said. According to him, the underlying industry fundamentals are sound and there is no end in sight to the ‘make-lunch-or-be-lunch’ ethos.

The emerging economies like India and China have so far been less affected by the financial market’s turbulence. In fact, the emerging and developing economies were shifting the global growth dynamics.

Chip industry in best possible shape
A forecast health warning is: IF the global economy collapses, it will take the chip market with it. However, Future Horizons feels that if there is going to be a global economic recession, the chip industry (but not all companies) is in the best shape possible to weather the ensuing storm.

The ASPs are an enigma wrapped up in riddle. The course of ASPs (like love) never runs smooth. Wobbles happen! ASPs are also the perennial (and least understood) industry wild card. ASPs are generally driven by new IC designs, and that takes time (sometimes three to four years). Post-2001, value recovery lost one generation (130nm impact). The ASP recovery ‘wobbled’ in 2007 (memory and MPU price wars). Barring a recession, Future Horizons forecasts that ASPs will recover in 2008 (it has already started).

12 percent growth likely
Future Horizons’ 2008 forecast summary and assumptions (as of May 2008) are — ‘12 percent’ growth — ’10 percent’ units / ‘2 percent’ ASP. There may be no global economic recession, although US/UK/Eurozone might wobble — which they are! No significant inventory correction will probably take place, but there are always Q4>Q1 adjustments, and there’s nothing special about that either.

There could be lower fab capacity expansion due to 2007/2008 capex slowdown, which is inevitable and irreversible. There is also a possibility of a more stable memory price erosion — which means, back to the learning vs. bleeding curve, and prices have since hardened. If the global economy holds, the 2H-08 growth will likely be strong. This, if the capacity, ASP and units are all pulling together, which is said to be happening.

Therefore, Penn feels it is too early to call for a (major) downward revision. Q1 08 was a lot stronger than conventional wisdom feared.

“That’s the rational analysis, but semiconductors aren’t rational. It could just as easily be another single digit growth year,” Penn added.

Danger signs to watch out for
So, what are the danger signs one should watch out for? These would be capacity — it is hard to see how this can spoil 2008, provided unit growth holds up, but there is a need to watch capex. Another factor is demand — the current IC unit demand is sustainable provided the economy holds up, so there is a need to watch the inventory.

Next comes the economy! The current outlook continues to be uncertain with risks all on the downside. ASPs are the key to recovery, but always the first line of defence. ASPs could still derail 2008, but the trends are encouraging.

What’s driving the market?
In semiconductor 7.0 — or the 7th decade of the transistor revolution, the same things, as always, are driving the market. These are: technology, legislation — energy saving/conservation and structural — the relentless analog to digital conversion. All of these are combining to do what the chip industry does best — enabling something that was previously impossible. Penn contends, “This industry has nowhere near run out of steam!”

New applications continue to drive the market, with automotive, industrial and medical, mobile phones, and PCs and servers, dominating. The PC market is dominating, but going nowhere fast. Mobile phones have become more interesting, but have conflicting priorities. The challenges are: how to protect the existing cost structure and subscriber base and how to add useful and affordable value-add services! Evidently, “chipset suppliers love the high end, market loves the low end.”

There is definitely an increasing automotive semiconductor content. A solid annual growth has been prediced (CAGR 2006-11) for vehicles — 5.5 percent, systems — 11.5 percent, and semiconductors — 13.3 percent. Some other new areas are motor control and energy, as well as lighting and photovoltaic, besides medical electronics. Robotics is yet another interesting area.

Key industry issues
It is clear that more chips per wafer equals less cost per chip and more transistors per die equals more functionality. Several billion transistors gives phenomenal design flexibility as well. Considering total ICs and MOS ICs, in the MOS capacity build out by technology node, there has been no change in volume ramp profile despite the hype.

As for the evolution of the technology node, definitely, 45nm is a revolutionary step from 65nm. In all likelihood, 32nm will be a natural evolutionary. However, Penn cautioned that 22nm would be another ‘difficult’ transition!

There is no doubt that 65nm will be tomorrow’s leading-edge workhorse, having the same basic Si gate/SiO2/MOSFET structure. Nevertheless, 45nm will herald a totally different structure — metal gate/high-k/thin FET/deep trench design, etc. Also, 45nm will herald a new way of system design.

Is fabless right?
Is Fablite a valid option? While there is nothing wrong with being fabless, people are just not sure whether the best starting point is being an IDM. Teamwork has to be perfectly orchestrated as competition is tough.

As for the market share dynamics, the top 10 companies (IDMs) have been losing share. Fabless share has been growing, but it is still relatively small.

Coming to the realities of the foundry market, TSMC’s lead is now unassailable. Were it an IDM, it would be No. 2, challenging Intel and passing Samsung. Moving more into design looks inevitable.

Finally, execution, and not technology, is everything! Execution has and will continue to make the difference. Applications (software) will play the role of the key differentiator as well, and it has value. Design is the means to an end, and not the end.

From the chip industry’s perspective, the electronics market was traditionally Japan, North America and Western Europe. It now encompasses the entire Asian rim, China, Eastern Europe and India. Far from maturing, the chip industry itself is still in its volatile, high-growth phase, with at least a further 20 years of strong growth in prospect. Penn said, “The underlying growth drivers for chips has never been better.”

Back to basics
We started with the need to get back to industry basics. We end in the same way! Stick to basics like:

* Don’t invest in low cost areas just because they are cheap — they have a habit of becoming high cost tomorrow, plus the hidden extras.
* Don’t make outsourcing decisions just because they are easy — especially if there’s no way back.
* Don’t make strategic cut-backs just to trim the bottom line — some decisions, e.g., R&D, take a long time to impact, then it’s too late.
* Stop looking for high volume/high value market niches — they don’t exist, need to learn how to compete
* Do show strong leadership
* Do have a long-term plan and stick with it — even if it negatively impacts ‘the next quarter’ balance sheet
* Do show a commitment and determination to succeed
* Do stay focused and resistant to external meddling
* Do execute ruthlessly — this is the key competitive differentiator)
* Do … just do it with passion — it’s the passion that makes the difference

Top 10 global semiconductor trends for 2008

December 20, 2007 1 comment

It is really difficult to stick your neck out and predict. That’s what makes the analyst’s jobs so difficult. Things happen and pass you by so quickly. For instance, as an example, who would have thought that Samsung would face a substantial blackout that would halt six chip production lines in a complex operated by the world’s largest flash memory producer?

Plans for the fab in India are now well under way. There have been questions like, do we need fabs? The year 2008 is the year of presidential elections and the Summer Olympics. Will we really see a recession in 2008? Here are some of the trends that are visible for 2008. Would love to hear from you.

1. Semiconductor firms may have to face a recession year in an election year

Yes, strange as it may sound, this just might happen! Concerns about consumer spending, caused by higher oil prices, mortgage crisis in the US and fears of a possible recession have made analysts more cautious, albeit optimistic. Analysts are wary of an impending recession in semiconductors during 2008. That, it should fall in the year of the US presidential elections makes it all the more intriguing. The nervousness is already showing in the slowing down of some markets.

2. DRAM market looks weak in 2008

Will DRAM prices rebound? Remains to be seen, although DRAMeXchange says that Taiwanese suppliers are likely to have their output to trim by 10-25 percent during February (Chinese New Year) as they usually plan for an average of 3-7 days of annual facility maintenance during this period. DRAMeXchange regards this as a possible catalyst for a price rebound in near term. Analysts haven’t helped either, with some saying DRAM will be on the slow side or even negative in H1-08.

3. NAND market will remain hot

You can bet, it will! Analysts remain upbeat for a positive NAND market in 2008. The reason being – new applications such as wireless USB, increase in cell phones memory capacities, higher content in portable media players, etc. We hope it is not a flash in the pan. There are rumors of another iPhone along the way!

4. Power will remain major issue

This isn’t going to change anytime soon! Power awareness is crucial for portable applications. It determines battery lifetime, and there’s an increased amount of computation involved as well. Power awareness is extremely crucial for high-performance applications. It determines cooling and energy costs. Many chip designs today are power limited and still require maximum performance.

5. EDA has to catch up

And fast! Analysts at a recent webcast hosted by Semiconductor International elaborated how the EDA industry was in a position of lag in the market. The DFM issue is increasingly becoming more complex. There is said to be a move to restrict the design rules that is in place now for 45nm. We are likely to see major changes in 32nm. That will have an impact on the EDA tools.

6. Need to solve the embedded dilemma

It is said that in 2007, the cost of designing or developing the embedded software for an SoC actually passed the cost of designing the SoC itself! We seem to be in the middle of a software crisis that is going to hit the entire electronics industry in the next five to six years.

Analysts are wary of an impending recession in semiconductors during 2008

7. Consolidation in the fab space

Some of the other older IDMs and fabs are said to be actually shutting down and going over to the foundries and process wafers for less than what they can do on their own. In this respect, we are seeing a lot of consolidation within the fab space. The mid-level players are consolidating. The customer base is clearly narrowing. The field is narrowing in 65nm and 45nm, and as we get to below 45nm, the field is going to get much, much narrower.

8. Capital equipment guys will continue to move to other markets

The best example, you can think of, is Applied Materials, which is into innovative equipment, service and software products for fabrication of chips, flat panel displays, solar photovoltaic cells, flexible electronics and energy efficient glass. Even the smaller guys are moving into LEDs or MEMS markets. That tells us what these companies are thinking about the semiconductors market.

9. Spend on capital equipment to drop

Gartner is expecting the long overdue capital spending correction in DRAM market to push the capital equipment market into contraction. Another slow year from foundry, along with concerns of US economic recession, adding to the downside. However, NAND spend should ramp up.

10. Mini fabs in developing countries

India has announced fab plans. There have also been talks of mega fabs and mini fabs elsewhere. There are going to be different types of fabs! With globalization, lot of countries may decide they want to have a fab. The market’s going to change.

However, bear in mind that the outlook on new fab starts appears weaker, as many companies have cut back on spending to wait for the market to improve. After a forecasted 8 percent YoY increase in fab construction spending in 2007, levels are likely to be flat in 2008.

Is the timing right for having fabs in India?

November 18, 2007 Comments off

Several majors have announced their Fab-lite strategies, and so, IDMs will become likely become fabless units of tomorrow. In this scenario, is the timing for setting up fabs in India right? What is the direction ahead for the Indian semiconductor industry?

Commenting on the upcoming fabs in India, Anil Gupta, MD, India Operations, ARM, said: “Yes, we do need the fabs to complete the ecosystem. The question is: Is the timing right? We have our own strengths. Why not capitalize on those?”

It is important to determine what India is doing as part of the global semiconductor industry. “What are we doing as an industry? Fabs are definitely a good idea. We also need to address things like — can we make products and more importantly, should we make products!”

Gupta pointed out Infineon, NXP, etc., had announced Fab-Lite strategies. Even Texas Instruments was moving to a Fab-Lite strategy. “IDMs are going to be the fabless units of today and tomorrow,” he added.

Coming back to the point of manufacturing products in India, he said: “We need to be able to conceptualize products for the mass market. Are we willing to take the risks? With services, the risks are significantly lesser. Companies are innovating on their service models. On the product side, India should do that as well. Maybe, we will do it too.”

As for the industry growth drivers, consumer applications would become even more atractive. “There are mobile phones, gaming applications and others, which will drive growth,” he said.

Yield management crucial
According to Gupta, yield management is crucial. While designers are well aware of yield management, the adequate tools are not yet there in place. The direct link
has yet to be established for implementing DFM/DFY technique.

Designers are always looking to prove how to improve yield. It is critical for designers to have access to the relevant information that would indicate that, say, some modification in design would lead to 10 percent increase in yield. Gupta said, “As much as we move to 45nm, to 32nm to 22nm and so on, the problems are going to become more complex and magnify.”

SOI addresses power, performance scaling issues
There is the silicon-on-insulator technology or SOI. SOI is said to improve power consumption, reduces leakage and allows better performance. Implementation of SOI technology is one of several manufacturing strategies employed to allow continued miniaturization of microlectronic devices.

ARM acquired SOISIC, a leading company in physical IP based on SOI technology last year. The move has enabled ARM to strengthen its physical IP portfolio by adding SOI technology.

“SOISIC’s niche is in developing SOI based IPs,” Gupta added. SOI addresses the power and performance scaling issues associated with traditional bulk CMOS processes as they migrate to ever-smaller geometries. “It is very clear that design starts for 45nm – 32nm – 22nm etc. will be very low. Each process geometry has to give returns,” Gupta added.

Role of IPs
So what’s the role of IP in the gameplan? ARM is trying to enable that on technology side with SOI and bulk CMOS. He noted: “You need building blocks to make things happen faster. From an IP perspective, analog IP is very, very closely tied to the process. In that respect, IP has a huge role to play.”

The semiconductor IP is said to be a $1.5 billion market. ARM currently has 30 percent share of that market.

From an Indian perspective, there are Indian companies who are building and also re-using IPs, as does ARM. However, ARM also has royalties for its IPs. On usage, Gupta said that physical IPs had greater challenges regarding re-use.

Mali55, Mali200 from India
Commenting on ARM’s India operations, Gupta said ARM India develops physical IPs, processors, etc. “There’s so much of verification and testing involved to make things work,” he added. ARM India currently has a workforce of 300+.

ARM India has done work on 65nm as well as physical IPs for 45nm. “We are also doing studies on 32nm,” added Gupta. ARM India released the Mali55 and Mali200 processors.

The ARM Mali200 graphics processor unit (GPU) delivers 3D graphics for next-generation mobile games on smartphones and other high-end portable devices. With a very small footprint, the ARM Mali55 processor brings rich 3D graphics capabilities to low-cost feature phones for the first time.

Besides these, a lot of software — embedded, drivers, stacks, etc., are being developed in India. ARM India also provides lot of support for various design implementations. “We have over 2,000 ARM certified professionals in Bangalore alone and over 7,000 in India,” he said.

ARM India has two other programs. Companies like HCL, Sasken, Mindtree, Wipro, etc., are ARM approved design centers (ADCs) or partners. “If we have any new product, we ensure that our partners become acquainted with those,” said Gupta. The other program is the ATC (training). Cranes Software is ARM’s approved ATC.

%d bloggers like this: