TSMC vs. Global Foundries - Let the War Begin

Just caught up with an August 3 article on TG Daily where Morris Chang compares himself (oddly enough) to Joseph Stalin! Seriously. Ah well, its all related to the ground breaking of the new Global Foundries Fab (GF) in New York and the signing of ST Micro as a customer. According to TG Daily:

San Francisco (CA) - The Chairman of Taiwan Semiconductor Manufacturing (TSMC) recently told reporters that he expected to triumph in a protracted and bloody war with GlobalFoundries. The septuagenarian compared himself to Stalin.

"We consider GlobalFoundries to be a formidable competitor," said Morris Chang, who was quoted by PC World. "I really think the battle will be a high casualty one. My job is to minimize the casualties on my side."

According to Chang, the construction of GF's 4.2 billion Fab 2 chip factory in upstate NY indicated a strategy of "total" committment. The opinionated Chang also compared GF's strategy to German attempts to hold the line at Stalingrad after being surrounded by Russian troops.

"Like Stalin, I have no doubt of the outcome," boasted General Chang.

GlobalFoundries spokesperson Jon Carvill responded to Chang's questionable analogy by reiterating GF's "total" commitment to fair competition.

"The groundbreaking in NY and the announcement of our newest customer, STMicro, were huge milestone for us and represents a long-term commitment to delivering the world's most advanced technologies in high-volume to the market," Carvill told TG Daily.

However, Carvill did concede that TSMC was a "strong" and "well established leader" in the foundry industry.

"We look forward to competing with them and offering a true alternative for those companies looking for the world's most advanced technology and manufacturing capabilitities," added Carvill.

Well well, we have talked about GF a bit in the past (see Global Foundries Challenges TSMC and Globalfoundries Getting in on the Game).

This fight will be an interesting one and one that TSMC has not really had in the past. They have dominated the foundry industry for a long time and have been the clear leaders. Although there are other competitors, it would seem that TSMC are way in the lead. This might be a bruising battle and one which the GF execs have had a lot of experience with. After all, they did challenge Intel (and lost a brutal price war) and now they are challenging the behemoth in the foundry industry.

Of course, GF does not have the experience in the pure play foundry industry and while they may have the manufacturing expertise (and a lot of money), and while the industries are related, they will still have to climb the curve. There is no escape. Their advantage (I would imagine) would be that many of TSMCs customers would love to reduce the supplier power that TSMC currently has in the market and for this reason alone, may outsource some of their production to GF. However, although TSMC does stand to lose a lot, I think the bigger fallout will come for the other (smaller players) in the foundry industry.

Time will tell but I still think Chang's Stalin analogy is odd! Why didn't he use Churchill?

TG Daily: "Intel's" TSMC declares war on "AMD's" Globalfoundries

Financial Crisis - TSMC Still Invests

EE Times reports that despite the economic downturn and the severe effect of the financial crisis on the high tech industry, TSMC is continuing to invest for the future while remaining cautious about the presence.

Amid one of the toughest periods in its illustrious history, Taiwan Semiconductor Manufacturing Co. Ltd. (TSMC) remains cautiously optimistic about the IC industry and vowed that it will continue to invest in R&D despite the downturn.

TSMC (Hsinchu, Taiwan) plans to hire more engineers. The world's largest foundry provider also reiterated plans to equip and ramp up its 40-nm fab lines this year. It is readying new and separate 3-D and CMOS image sensor technologies. And it is also planning to move the IC-equipment in its R&D fab for the 22-nm node.

Rick Tsai, president and chief executive of TSMC, reiterated industry reports that the silicon foundry giant is seeing new order activity, but he also warned that there are still challenges ahead in the market.

Among those challenges include the overall economy, product demand and margin pressures. "This recession is bad," Tsai said at TSMC's Technology Symposium here. "This is a difficult time for all of us."

Indeed, it has been a humbling time for TSMC. After strong growth in the first three quarters of 2008, TSMC's business fell off the cliff in the fourth quarter of last year.

As a result, the company is expected to report a loss in Q1. It also recently cut about 200 jobs, implemented furloughs and slowed its wafer starts.

Now, there are some positive signs for the company and the overall industry. Inventories are low. Activity in China is picking up. "We are seeing what we call rush orders," Tsai said.

Still, the overall IC market is expected to fall in 2009. "We will see a dip in 2009," he said. "We will see moderate growth in 2010." [...More]

We have argued more times than not in this blog that despite the economic downturn, companies must continue to invest in their future and develop strategic plans that will pull ready them for when the slump is over.

SMIC Posts Big Loss

It seems the foundry industry is getting very very crowded with the emergence of Global Foundries. The industry will be forced to restructure itself and some of the weaker competitors will face bigger challenges. Of course most people (including us) have focused on the obvious rivalry between GF and TSMC but the other foundries will also be caught in the struggle. Digitimes reports China's largest foundry had the biggest drop in revenues in five years.

Semiconductor Manufacturing International Corporation (SMIC), China's largest silicon wafer foundry, experienced its largest net operating loss over the past five years in 2008. The company posted a loss of US$440 million, compared to an operating loss of US$19 million in 2007.

SMIC saw its revenues slide 12.7% to US$1.35 billion in 2008, which the foundry attributed to capacity adjustment at its Beijing fabs as well as the world economic meltdown. [...more]

Well they are still facing a tough industry environment and with addition of a new competitors, the smaller foundries might also have a big fight on their hands.

Look forward to your comments.

Global Foundries Challenges TSMC

Well the challenge has arrived. Global Foundries are pushing 28nm technology.

An alliance of technology companies, which includes IBM, Samsung, Chartered, Infineon, STMicroelectronics and the spin-off of AMD's manufacturing operations - GLOBALFOUNDRIES (GF) - has announced the joint development of 28 nanometer semiconductor manufacturing process technology.

The significance of a shrink in the manufacturing process technology (the most recent CPUs are made using a 45nm process) is that it allows either an increase in outright processing performance or a reduction in the size and power required to achieve the same performance.

With GF being the sole manufacturer of AMD processors, this announcement is being viewed by many as a shot across the bows of Intel.

There's no doubt that any progress in semiconductor research from competitors to Intel is significant, but GF is now a supposedly independent semiconductor foundry and, as such, has its sights set on its own competitors, the biggest of which is TSMC (Taiwan Semiconductor Manufacturing Company).

For GF to be a success, it has to appeal to a lot more companies than AMD. The development of a cross-over node at 28nm (full nodes either side are 32nm and 22nm) is being positioned as significant because it may create a differentiator between GF and TSMC. If GF can use that differentiator to take customers away from TSMC, that would represent a significant achievement. [...more]

But TSMC are already expecting to have 28-nm production in Q1 2010 and are already working with customers (see here) to develop 28-nm chips. Maybe the half-node positioning at 28-nm will provide Globalfoundries with some temporary advantage (TSMC at 28-nm will be a full-node process) but they still have to sell it. The struggle will be fun to watch but as one commentor on our blog suggested, GF really have to completely dissascciate themselves from AMD to gain trust in the market. We will wait to see if that happens.

Look forward to your comments as always.

Globalfoundries Getting in on the Game

Competition in the pure play foundry industry is about to heat up. ars technica has an interesting blog article on AMD's spinoff company Globalfoundries who seem to be on thr prowl for TSMC and UMC customers. ars technica writes:

It's launched, online, and the now-independent Globalfoundries is searching for its non-AMD customers. CEO Doug Grose is reportedly visiting Taiwan to seek relationships with companies that may be currently contracting with TSMC or UMC for their semiconductor foundry needs. If true, this would raise questions regarding the future of AMD's relationship with TSMC. That company currently fabricates Radeon processors for Advanced Micro Devices, and while Globalfoundries and AMD are now separate entities, they are separate entities that remain joined at the hip. Globalfoundries will probably take over fabbing ATI Radeon processors at some point, but is not yet believed to have the bulk silicon production in place to do so.

So what are your thoughts? Will Globalfoundries actions in Taiwan threaten the relationship between AMD and TSMC? Will Globalfoundries be able to compete effectively? Are TSMC and UMC in a world of trouble? I will be interested to read your comments. In the meantime, I look forward to see how the pure-play foundry industry landscape changes. (Remember last week TSMC and Intel signed an agreement). The future is always exciting, especially in this industry

ars technica: Globalfoundries on the prowl for non-AMD customers

Intel and TSMC Strategic Partnership

Intel and TSMC have struck a deal to co-manufacture Atom system-on-chips. According to PC Mag:

Intel and foundry Taiwan Semiconductor Manufacturing Co. have struck a deal to allow customers to design their own Atom system-on-a-chip processors and manufacture them at TSMC.

Intel is not outsourcing the Atom processor, as this reporter speculated on Friday. Customers who wish to buy standalone Atom chips will buy them from Intel, and Intel has not altered its Atom roadmap or production.

Intel, however, has also made the Atom a cornerstone of a system-on-a-chip strategy, such as the "Moorestown" and "Lincroft" for the mobile Internet device market. Now, an Intel customer will be able to use TSMC's process flow, tools, and intellectual property (both from TSMC and its partners) to create their own Atom-based system-on-a-chip products, which will be sold under the Intel brand.

"It's enabling Atom on TSMC,"said Anand Chandrasekher, general manager of Intel's Ultra Mobility division, during a conference call Monday morning. "It will allow TSMC to go after new market segments and allow Intel and TSMC to go after new market segments together."

This deal will inevitably be more beneficial for TSMC but it will also enable Intel to penetrate other market segments more easily and one would guess enable system-on-chip designers to leverage some of the advantages of the Intel Atom processor for their own designs. Its an interesting partnership to say the least.

PC Mag: Intel, TSMC Strike Atom Design, Foundry Deal

Morris Chang Predicts Three Year Recovery

Interview in the Wall Street Journal, Morris Chang anticipates the semiconductor industry will take three years to recover to 2008 sales levels. The pounding the semiconductor industry has taken over the past few months due to the financial crisis has forced semiconductor sales to drop dramatically. According to the WSJ:

The global semiconductor industry has yet to hit bottom and likely won't recover fully from the current downturn for another three years, according to one of the industry's senior figures, Taiwan Semiconductor Manufacturing Co. Chairman Morris Chang.

In an interview Friday, Mr. Chang said the industry, which has been hard hit by the global economic slowdown, was "pretty close" to the bottom. But he said that a return to the sales volumes that the industry had before the current slump would be extremely slow. "You get a precipitous drop and a very slow rise," he said.

"I think it will be 2012 before the total revenue of the semiconductor industry gets back to the '08 level," said Mr. Chang, who founded Taiwan Semiconductor in 1987.

The WSJ continues:

Mr. Chang also said he expects industry consolidation to pick up. He predicted a continued decline for companies that make both consumer products and semiconductors. Of the companies that still do both, he said that only Samsung Electronics Co. and Intel Corp. are in a strong position. He said others would likely divest themselves of their semiconductor operations, leaving production to companies like TSMC, which manufactures semiconductors on a contract basis for other companies.

Mr. Chang has worked in this industry for a long time and his thoughts do carry some weight. It will be interesting to see how this all pans out over time, especially the consolidation of the foundry industry. What will be more interesting will be to see how the new foundry company being spun off will survive. The foundry business is a tough business to play and one can only imagine the perils of entering the industry at this point in the economic cycle.

Wall Street Journal: Executive Sees Chip Industry Recovery in Three Years

TSMC Stays on Top

An interesting article on EE Times shows Taiwan's foundry giant TSMC continues to grow and gain market share. According to EE Times:

Amid the downturn, which up and which is seeing red? An analysis from FBR Capital Markets found that Taiwan Semiconductor Manufacturing Co. Ltd is looking up, but the rest are down.

According to FBR, recent checks suggest that the average foundry wafer shipment in Q 09 will be down by nearly 40 percent q-on-q (versus the historical average of down 5percent), following a 30 percent decline in Q4 08 (versus a historical average of up 5 percent). "Our current assessment for Q2 09 is for the average foundry wafer shipment to be flat to up 2 percent quarter-over-quarter (versus a historical average of up 10 percent) with the exception of TSMC, where we expect wafer shipment to be up 8 percent quarter-over-quarter."

The firm believes that TSMC has remained on track with an overall wafer shipment decline of 35 percent to 40 percent q-on-q in Q1 09, though Q2 09 wafer shipment forecasts have recently been revised up from up 3 percent to 5 percent q-on-q to up 8 percent q-on-q.

"Additionally, we believe that TSMC has gained incremental market share at AMD at the expense of Chartered Semi. Foundry checks indicate TSMC is well positioned to continue to outperform and gain market share, while demonstrating superior cost/margin structure," FBR said.

I have personally argued for a while now (not on this blog) that the economic downturn is going to force bad companies to close and the well managed companies that pay attention to scale, cost, efficiency, market share and debt/equity ratios will suffer but survive. TSMC is one of those companies that will continue to survive. It does help that they have significant market share but there capabilities and continued progress in developing and maintaining their know-how should see them through. Of course they will also suffer in the downturn, but their continued success does not surprise me. But, as a word of caution, we can never really know what tomorrow brings!

EE Times: Foundry watch: TSMC stays on top

Foundry Utilization Rates Set to Drop

The financial crisis is set to continue to depress the foundry industry into Q1 2009. According to one analyst, utilization rates in the foundries are set to drop by a significant amount. Solid State Technologies reports:

Wafer shipments at the world's top two foundries, TSMC and UMC, are set to plunge further than anticipated in 4Q, but the picture for 1Q is even uglier with "historic lows" looming for utilizations, according to an analyst report.

Wafer shipments sunk 30% in 4Q vs. 3Q vs. a historical average of 5%-8% growth, and FBR Research's Mehdi Hosseini says checks indicate they'll keep dropping another 20% in 1Q09, vs. a typical -5% decline, with weak demand seen in "all customers across the board," he writes. Foundry customers are seeing inventories decline, suggesting the sluggish shipments are due to weak end-demand. Look for some eye-popping floors in foundry capacity utilization, he warns -- "reaching 50% and below levels," which will further depress equipment spending.

It is no secret that part of Taiwan's success in the pure play foundry model has been there ability to maintain their high-utilization rates. Declining utilization will definitely affect their competitive advantage over the short term. Silicon foundry's are not cheap to run and any significant decrease in orders and production will greatly impact the cost of production and naturally effect gross margin these companies can expect to earn. This is certainly going to impact their abilities to develop and introduce new technologies into the manufacturing process and may even provide a space for other competitors to gain an edge over these foundries although one should think that the low utilization is spread accross the entire industry and not only affecting TSMC and UMC.

Wafer News: Analyst: Foundries face "historical lows" in utilization

More Pressure from the Financial Markets

Yesterday in Madness in the Markets we looked at the effect of the financial meltdown on some tech sector companies. Following this theme today we read on the China Economic News (CENS) that spending on semiconductor manufacturing equipment is predicted to be in decline. According to CENS:

Semiconductor Equipment and Materials International (SEMI) recently cut its 2009 forecast of global capital expenditure on semiconductor manufacturing equipment to somewhere between a 5% growth and a 10% decline, sharply down from a 20% surge it previously forecasted.

The trade organization, which represents suppliers of semiconductor equipment and materials and suppliers of liquid-crystal display equipment and materials, ascribed the bearish forecast mostly to the worsening global financial crisis.

Some equipment suppliers pointed out that the market is now anemic. Taiwan Semiconductor Manufacturing Co. (TSMC), the world`s No.1 silicon foundry, is reported to further cut on expenditure plan for next year while No.2 supplier United Microelectronics Corp. (UMC) is said to spend US$400 million, compared with this year`s US$600 million outlay.

Total 2009 expenditure by Taiwan`s top four dynamic random access memory chipmakers-PowerChip Semiconductor Corp. (PSC), ProMOS Technologies Inc., Nanya Technology Corp. and Inotera Memories Inc.-is estimated to be only half the NT$88 billion (US$2.7 billion at US$1:NT$32) they spend this year. Weak market has sent spot market price of 1-gigabyte DDR2 chip slumping to only US$1.2 apiece.

SEMI estimated the financial meltdown to force global spending on equipment down by 20% throughout this year. Applied Materials, the world`s No.1 supplier of semiconductor manufacturing equipment and materials, even announced a drastic 40% drop.

Such downbeat mood has infected fabless houses: Realtek Semiconductor Co., Ltd.`s vice president, J.S. Chen, estimated the downturn would last at least six to nine months.

Industry watchers pointed out that, unlike past market downturns, this downtrend is across-the-board, giving almost no sector a chance to bring about a boom cycle. To weather such hardship, chipmakers choose to hold more cash.

I don't think any industry is immune to the meltdown. The semiconductor manufacturing is an expensive game and holding on to cash is an important part of the equation. Highly-geared companies in this environment with uncertain or declining cash-flows may be in serious trouble. Semiconductor manufacturers will then be loath to spend on expensive fabs.

As if to reiterate the effect of the financial meltdown on different companies,Winbond revenues decline by 40%. According to Digitimes:

Winbond Electronics reported September revenues of NT$1.677 billion, a decrease of approximately 9.25% compared with NT$1.848 billion in the previous month. September's sales were also down more than 40% from the same period one year earlier.

Accumulated revenues for January to September of 2008 were NT$18.05 billion, down 29% from one year earlier.

September business was negatively affected by a soft DRAM pricing. Under a weakening macroeconomic environment and in order reduce losses, Winbond continues shrinking technology, driving cost reduction and adjusting production capacity to niche products.

Winbond are struggling but at least they are proactively trying to reposition themselves by focusing on cost reductions and targeting niche product markets.

CENS: SEMI Cuts 2009 Forecast of Semiconductor Equipment Capex
Digitimes: Winbond September sales down 40% on year

AMD Split and Implications for TSMC

I am scratching my head here a bit because I just read an article on internetnews.com that looks at AMD's foundry spin-off (The Foundry Company) competitive targeting of TSMC. The reason I am scratching my head is that they claim AMD will have a technological advantage over TSMC because they already have 45 nm processing technologies in place. According to them (emphasis added):

With a long-term goal of being a player in the semiconductor foundry space, TFC now has its sights set on TSMC and other established players. But it won't have the capacity to go head-to-head, notes Spooner. While TSMC has 11 fabs, AMD had only its Dresden facility and its planned New York facility, which is years away from completion.

But TFC will also have an advantage of being much more cutting-edge in its manufacturing capabilities than TSMC, putting it closer to Intel. TSMC is just reaching 55 nanometer die sizes, while AMD, thanks in part to a technology alliance with IBM, is at 45nm and heading to 32nm.

The Foundry Company will join IBM's joint development alliance, a group of leading semiconductor companies collaborating on next-generation silicon technologies, with the ultimate aim of reducing die sizes to 22nm.

This means mean TFC can go gunning for TSMC customers. TSMC has an impressive collection of customers, including Broadcom, Conexant, Marvell, NVIDIA, VIA and ATI, which AMD now owns.

"The Foundry Company will be aimed at the very high end of the market, because they will offer very cutting-edge technologies and will be able to come to market before TSMC," Spooner told InternetNews.com. "I'm not sure how much revenue they can generate going forward. It all depends on how quickly they bring on partners."

This will also let TFC invest more heavily in the fabs than AMD could have all by itself, and it can make more money by making chips for other firms than AMD. AMD's fabs have some problems, most notably they don't use 300 millimeter wafers yet, they still use 200 mm.

A larger wafer means more chips can be made at once, and thus reduce manufacturing costs. Spooner said that with the backing TFC now has, it can make that transition.

First up, TSMC does have a 45 nm fab. All one needs to do is check their website here and you will see the following graphic:

Notice they do have at least one 45 nm fab with an additional two in the works and a plan for three 32 nm fabs. They also seem to have plans for 28 nm processing technology by 2010. TSMC is very competitive in the foundry business and should not be taken lightly.

There is also an issue of long term contracts and established partnerships with TSMC. TSMC also has a great deal of knowledge of how to be competitive in the pure-play foundry industry. AMD spin-off The Foundry Company (TFC) may have big ideas, but it will not be as easy for them. They are going up against a smart competitor with large market share. Remember the last time they did that against Intel they lost.

At any rate, I guess TFC knows more about TSMC's capabilities than either me or the author of that article and I am pretty sure they will not enter this market naively. If they do, they will be in trouble.

Internetnews.com: AMD Dumps Fabrication Plants

AMD Loses Their Manhood

In July in Is AMD Going to Spin-Off their Manufacturing we noted foundry industry observers were commenting AMD were going to have to spin-off their fabs to stay viable. Later we noted in AMD's Asset Smart Strategy: What is it? AMD had adopted an asset-smart strategy. Although this strategy was not clear it was predicted that AMD meant they would be dumping their fabs. Well, its come to pass. AMD will no longer manufacture their own chips. According to the New York Times (NYT) AMD are going to split into two companies. One will focus on chip design and the other on manufacturing. According to the NYT:

Advanced Micro Devices said Tuesday that it would split into two companies — one focused on designing microprocessors and the other on the costly business of manufacturing them — in a drastic effort to maintain its position as the only real rival to Intel.

In addition, the company said two Abu Dhabi investment firms would inject at least $6 billion into the two firms, mostly to finance a new chip factory that A.M.D. planned to build near Albany, N.Y., and to upgrade one of the company’s existing plants in Dresden, Germany.

A.M.D., based in Sunnyvale, Calif., makes graphics, computer and server processors. It will own 44.4 percent of the new entity, which has been temporarily named the Foundry Company, a reference to the technical term for a chip factory. The Advanced Technology Investment Company will own the rest.

Advanced Technology, which was formed by the Abu Dhabi government, has promised to put up $2.1 billion immediately and contribute $3.6 billion to $6 billion more to build or upgrade chip fabrication plants, also known as fabs. A.M.D. said the two companies would share voting control equally.

The Mubadala Development Company, an Abu Dhabi company that bought 8 percent of A.M.D. in November, will pay $314 million for 58 million newly issued shares, increasing its stake in the presplit company to 19.3 percent. It will also get warrants to buy 30 million shares. A.M.D. stock closed Monday at $4.23 a share, down 30 cents.

“We generally believe this deal is a game changer for the industry,” said Khaldoon Al Mubarak, chief executive of Mubadala. “It’s bold, and I think it’s smart.”

Coming up with the billions of dollars needed to construct each new chip plant has proved to be a huge drain on A.M.D., the perennial No. 2 to Intel in the market for microprocessors, the powerful chips that control the functions of personal computers and the larger corporate machines known as servers. As of June, A.M.D. reported that it had $5.3 billion in debt and just $1.6 billion in cash.

With the constant need to devise smaller, faster, more energy-efficient chips to keep up with Intel, A.M.D. was forced to turn to outside help.

Well this turn of events was certainly not unexpected. AMD have been smashed around for the better part of three years in a ruthless price war with Intel. They are on the ropes and there is nowhere to go except to be broken up and allow the latent value in either firm to be unlocked somehow.

This will however have an impact on the global semiconductor industry. A new foundry provider in these dark times will increase the competitive environment in the pure play foundry business and smaller competitors may feel the pinch of it. I suspect TSMC, the world's pre-eminent pure-play foundry, will not be effected initially but they certainly cannot ignore the emergence of a new player in the game.

So why is this piece titled "AMD Loses Their Manhood"? The NYT answers (emphasis added):

The split, which has been in the works for more than a year, did not come easily to A.M.D. According to company lore, A.M.D.’s co-founder and longtime chief executive, W. J. Sanders III, known as Jerry, once remarked that “real men have fabs.”

Under the deal proposed by A.M.D., the company would retain many of the traditional benefits of fabs, since part of Foundry will be dedicated to serving A.M.D. and will remain in close communication with the company’s engineers.

“We feel like we’re still pretty manly at A.M.D.,” Mr. Meyer said. Noting that Mr. Sanders made his quip over a decade ago, he added, “Frankly, the math has changed.”

Welcome to a whole new semiconductor landscape, the future will be interesting.

New York Times: A.M.D. to Split Into Two Operations

Morris Chang Looks to the Future

Morris Chang, founder of Taiwan Semiconductor Manufacturing Corporation (TSMC) believes the future of the semiconductor industry, and speficially the manufacturing side of the industry is in the hands of three global giants: Intel, Samsung and TSMC. The Statesman commented on a speech by Morris Chang saying:

As Morris Chang looks to the future of the semiconductor industry, he sees technical and economic challenges that will strain chip makers' bankrolls and their intestinal fortitude.

In a decade or so, he expects to see only three companies still devoting major resources to pushing chip manufacturing technology dramatically forward.

The final three, he says, will be Intel Corp., the long-standing kingpin of the industry, South Korea's Samsung Electronics Co., the biggest maker of memory chips, and Taiwan Semiconductor Manufacturing Co.

The Statesman continues saying:

Chang believes the chip industry can gradually push ahead to the next two or three generations of advanced chip-making technology before really big challenges start to appear.

When that happens, the number of chip makers that will keep spending heavily on research will start to shrink.

The reason only three will remain is the prohibitive cost of the technologies and the plants. One plant costs in the region of US$5 billion. The Statesman wrote:

State-of-the-art chip factories now cost as much as $5 billion to build, and they require enormous production runs to operate efficiently. Fabless smaller companies would rather let TSMC take care of the chip manufacturing while they concentrate their efforts and their money on innovative design and aggressive marketing. And so would increasing numbers of large companies.

It is hard to see anyone else from capturing market share in the future from these three companies. AMD tried to compete with Intel and got hammered into the ground during an extensive price war over the past few years that has left the company teetering on the brink.

Digitimes recently reported that TSMC had "94% of all profits generated by four wafer foundries in Taiwan." According to Digitimes:

TSMC "revenues were up 25% on year to NT$170.8 billion (US$5.39 billion) in the first half, whereas net profits increased 28% on year to NT$56.9 billion during the same period.

United Microelectronics Corporation (UMC) saw its net profits decline 59% on year to NT$2.6 billion due to falling non-operating income, losses incurred from foreign exchange transactions and declining financial asset values, the company said.

That is not to say the fabless design model is without its threats and dangers. One recent analyst suggested companies like Nvidia may have to reavluate their relationship with TSMC as TSMC is unable to guarantee them production runs. EETimes comments:

Nvidia Corp. is not getting the 55-nm capacity it needs from silicon foundry giant Taiwan Semiconductor Manufacturing Co. (TSMC), a problem likely to worsen as the graphics chip maker moves to the 40-nm node, according to Doug Freedman, an analyst at American Technology Research.

In a report published Thursday (Sept. 11), Freedman wrote that the "fabless business model seems to be getting stressed out" and that the Nvidia-TSMC relationship "sounds problematic" and "needs to evolve."

In order to minimize its own risk, TSMC is building less capacity as it ramps new technology nodes, according to Freedman. TSMC doesn't get enough visibility from its smaller customers such as Xilinx Inc., Altera Corp. and Broadcom Corp., Freedman wrote, so the foundry cannot take the risk of building new capacity to support larger customers. Nvidia is TSMC's largest customer, he noted.

TSMC's board of directors last month approved a $795 million capital spending plan that includes a push into 45-/40-nm CMOS processes and MEMS.

Maintaining high capacity utilization and managing the risk of not receiving sufficient orders for maximum utilization rates is a delicate balancing act for TSMC. One which they will have to manage.

Maybe Morris Chang will be proved right, then again maybe not. We will have to wait and see in 2018 who is still around.

The Statesman: A chip prediction: In the end, three will dominate R&D
Digitimes: TSMC tops Taiwan wafer foundries in revenues and profits in 1H08

TSMC Increases Their CAPEX

EE Times writes TSMC has decided to expand their CAPEX for to develop their 45 nm and 40 nm processing technologies. Some of the CAPEX will also be used to upgrade their equipment. EE Times writes:

The board of Taiwan Semiconductor Manufacturing Co. Ltd. (TSMC) has approved a $795 million capital spending plan that includes a push into 45-/40-nm CMOS processes and MEMS.

The spending is part of TSMC's previously-announced capital expenditure budget for 2008. For year 2008, total capital expenditures for TSMC is expected to be around $1.8 billion, compared with $2.6 billion spent in 2007.

On the leading-edge front, silicon foundry giant TSMC said it has approved the appropriations of $687.60 million to expand its 45- and 40-nm manufacturing capacity within its 300-mm fabs. TSMC recently rolled out these processes.

The company has also approved capital appropriations of $107.40 million in 200-mm fab equipment. It will upgrade a portion of its 0.18-micron logic process capacity to 0.11-micron CMOS image sensor technology, 0.11-micron logic, 0.13-micron high voltage, and 0.18-micron RF.

EE Times also notes TSMC's expansion into the Microelectromechanical System (MEMS) fabrication (we previously commented on this in Taiwan Manufacturers Move into MEMS market). According to EE Times:

In addition, the appropriation will also be used to upgrade a portion of its 0.35-micron logic process capacity to MEMS processes.

TSMC (Hsinchu) has been involved in MEMS foundry production for some time, but the company is expanding its efforts in the arena--and for good reason: MEMS is growing at an annual rate of 13 percent, according to the company.

Within its fabs, the company is devising several MEMS products on a foundry basis for customers, such as inkjet devices, sensors, RF MEMS and displays. It is developing several processes in the arena, such as bulk MEMS, surface MEMS and a CMOS-MEMS integration technology.

Unsurprisingly UMC, TSMC's main competitor, is also jumping on the MEMS bandwagon.

EE Times: TSMC hikes capex for 40-nm, MEMS

AMD's Foundry Business

As we have noted on and off over the past AMD seems to be becoming an asset-lite or fabless design house (see Is AMD Going to Spin-Off their Manufacturing). AMD have denied they are spinning off their manufacturing and insist they will keep their fabs. The problem is if they keep their fabs they will be less competitive against Intel as the fabs suck up a lot of money. The relevance of this to Taiwan tech firms is that if they do spin-off their manufacturing, the chances are they will use Taiwan's pure-play foundries to manufacture their products. An excellent commentary on EE Times highlights the difficult choices facing AMD.

EE Times says:

In the future, AMD is supposedly planning to split the company into two parts. It will supposedly spin out its manufacturing unit, turning AMD into more or less a design house.

The manufacturing spinoff may take ownership of AMD's fabs, especially its leading-edge plant in Dresden, Germany. Reports have surfaced in the media that AMD will also have its parts made on a foundry basis by Chartered, TSMC and possibly United Microelectronics Corp.

Sources say otherwise. TSMC is expected to get the lion's share of AMD's foundry business--leaving Chartered on the outside looking in, according to one source. Chartered may make some of AMD's products, but the many of AMD's key parts have already migrated to TSMC's fabs, according to the source.

In any case, the strategy is risky. There is little or no evidence that a foundry can keep up in the high-volume processor game against Intel. Foundries can make processors, but they have typically worked with small, third-tier suppliers over the years. Competing with Intel is a different story.

On second thought, there's a good change the strategy will succeed. By the time AMD makes its asset lite announcement--and based on the company's recent execution--the processor vendor could lose what's left of its status and become a second- or third-tier player.

Maybe not a third-tier player, but one that is fighting to stay relevant.

I find it interesting the commentary argues the foundries won't be able to maintain their high-volume processor manufacturing. I don't think the pure-play foundries have ever competed with Intel directly as the business models are different. I imagine the problem here would be the utilization rates of the fabs. To compete evenly with Intel a significant amount of the fabs processing capacity will have to be used to manufacture the processors. This will have a negative effect on the pure-play foundry's other customers. Its an interesting dilemma that they will have to solve themselves.

EE Times: Commentary: Who won AMD's foundry business?

More Foundry News

Tech-On has an insightful article on the development of TSMC's 300 mm (12-inch) fabs in Hsinchu, Taiwan. Tech-On writes:

Taiwan Semiconductor Manufacturing Co Ltd (TSMC) explained the current condition of its 300mm plants as well as its business results for the second quarter of 2008 at a press conference in Tokyo August 1, 2008.

Fab 12 and Fab 14 in Hsinchu are the company's main 300mm plants. The production capacities of the plants are being enhanced to make them "Giga Fabs" with a capacity of more than 100,000 300mm wafers per month. To meet strong demand for the advanced processes of the 65nm generation and beyond, the company has been making capital investments ahead of schedule.

Tech-On further notes the convergence of the foundries with the fabless design houses suggesting that as technology advances, the foundry is becoming more involved in the design of the actual chips. Tech-On writes:

TSMC operates a silicon foundry business with enormous production capacity, but its relationships with users, fabless companies, are changing recently. Until recently, there was a distinction between the roles and users were responsible for designs, while TSMC was responsible for process development and manufacturing.

However, since several years ago, the boundary between their roles has not been as clear as it used to be. I many cases, TSMC works closely with a user from the initial stages of process development and collaborates with a user till the last stage of production, not to mention the designing process.

Following on from our story Fabs to Stay in Taiwan , Digitimes notes that while senior executives at both TSMC and UMC have welcomed the relaxation of laws regarding foundry investments in China, both companies are indifferent to investments in China. Digitimes notes:

When semiconductor demand takes off in the local China market, that will be a judging factor for making the move, said Rick Tsai, CEO of TSMC. While saying that TSMC welcomes an open policy, Tsai added that investing in 12-inch wafer fabrication in China would require taking economic and administrative issues into account.

Morris Chang, chairman of TSMC, indicated that TSMC will focus on expanding its 12-inch wafer capacity at its Hsinchu, Taiwan base. Tsai noted that TSMC is consistently growing its capacity at its Shanghai 8-inch fab (Fab 10). He highlighted that customer base at Fab 10 is broad and he believes that Fab 10 will be a direct beneficiary of a booming semiconductor industry in China. But both of the executives said there seems no urgency to invest in 12-inch wafer fabrication in China.

Shih-Wei Sun, CEO of UMC, said the company has no special comments about potential 12-inch wafer fab investment in China. Company CFO Chi-Tung Liu added that UMC has no plans yet to construct a 12-inch fab in China, as the company will focus on expansion at Fab 12B in the near term. Monthly capacity at Fab 12B could reach 50,000 wafers in the future, he added. UMC has taken three years to ramp capacity at Fab 12A to a 30,000 wafer level.

Digitimes also notes IC Insights has shown TSMC had the largest growth of the top 20 semiconductor companies. Digitimes says:

Recent research from IC Insights uncovered a big shakeup in its global top 20 semiconductor supplier rankings, with many of the major DRAM and flash suppliers (e.g., Qimonda, Elpida, Spansion, Powerchip, Nanya, etc.) no longer part of the top 20 ranking. Although the top four ranked companies remained the same from last year in the rankings – Intel, Samsung, Texas Instruments, Toshiba – it was fifth ranked Taiwan Semiconductor Manufacturing Company (TSMC) that showed the strongest growth among the top 20, with its sales up 35% on year.

Tech-On: TSMC Discloses Under-construction 300mm 'Giga Fabs'
Digitimes: Relaxed restrictions welcomed in Taiwan, but indifference remains about actually building 12-inch wafer fabs in China
Digitimes: TSMC leads growth among top 20 semiconductor suppliers in 1H 2008, says IC Insights
IC Insights: Shakeups Rock 1H08 Top 20 Semiconductor Supplier Ranking

Fabs to Stay in Taiwan

In recent weeks we have been following the proposed new legislation in Taiwan that will allow advanced 12" semiconductor fabs to be moved accross the straits to China. In Mr. Ma please don't move the fabs to China and China Strategy for Fabless Chip Designers we argued Taiwanese firms should be careful moving their most advanced technologies to China because of the lack of protection for intellectual property rights. We noted this lack of proctection in Hon Hai Fights in Shenzhen where Hon Hai Precision Technologies have sued their Chinese competitor but deliberate delays in the courts have resulted in their competitor being able to strengthen their business position.

In No 10G AUO Plant in China we conceded that any investment decision into China will ultimately be a business decision and not a political one. Although Taiwanese legislation may change, the business leaders still need to do their own analysis to decide if it is worth the investment. Investing in fabs is afterall an expensive affair. Intel's 12" Dalian fab for example is costing in the region of US$2.5 billion. We were therefore not surprised to read in PC World that the two biggest pure-play foundries in Taiwan, TSMC and UMC, have no immediate plans to invest in China. PC World writes:

The Taiwan government is planning major changes to regulations governing chip-related investments in China, but the island's companies are in no hurry to build new factories on the mainland.

"Currently, we do not have a fab expansion plan in China," said Rick Tsai, CEO of Taiwan Semiconductor Manufacturing (TSMC) at the company's second-quarter investors' conference on Thursday. The company's current Shanghai facility will reach its maximum capacity by the end of the third quarter and TSMC owns enough land at the Shanghai site to build a few more fabrication plants, or fabs there.

But the world's largest contract chip maker sees worsening business conditions on the horizon, caused by high energy prices and rising materials costs.

Other chip makers have also noted problems.

Taiwan's DRAM makers, for example, continue to post losses in a tough market for their chips. DRAM prices have fallen below the cost of production due to a chip glut, leading to the losses. Taiwanese DRAM makers have all cut back on their expansion plans.

TSMC's biggest rival in the contract chip making business, United Microelectronics (UMC), doesn't plan to build any new factories in China soon, either.

UMC is focused on legally taking ownership of a 15 percent stake in China's He Jian Technology, a contract chip maker UMC executives have admitted to aiding in its start-up phase. But UMC and company executives say they broke no Taiwanese laws by helping the Chinese company. He Jian is UMC's partner in China.

"Aside from the He Jian stake, we have no other plans to invest in China," said Sun Shih-wei, CEO of UMC, during the company's second quarter investors' conference on Wednesday.

A number of global companies have been attracted to China for its huge market, low-cost labor and incentives for chip related investment, such as construction subsidies, low-cost land in special technology park zones, and tax breaks.

Last year, Intel, the world's largest chip maker, announced plans to build a US$2.5 billion chip fab in Dalian, on China's northeastern coast.

What will be interesting to see is what incentives the Chinese government will use to try to attract the fabs to China. Incentives offered to Intel made building the fab in China US$1 billion cheaper than the states. The Chinese government aggressively pursues Taiwanese investments on the mainland as we noted in Attracting Taiwan. No doubt over the next few years incentives will be offered, but my guess is the Taiwanese businessman are fully aware of the risks attached with moving these investments to China and will therefore be extra cautious.

PC World: China May Have to Wait for Chip Investment From Taiwan

The Foundry Industry -- A Recent History

In TSMC sees good Q2 but Expects Slowdown we noted some of the successes of TSMC and how TSMC project only a 4% growth over the next quarter. The Financial Times, in response to TSMC releasing their financials, has an excellent article on the last downturn in the industry when most industry experts believed TSMC would battle to survive. The FT argues the situation is now reversed and TSMC is very well placed to struggle through the next downturn. According to the FT:

TSMC, the world’s largest contract chipmaker, predicted flat sales and a slight squeeze on margins in the current quarter. Nevertheless, many analysts are keeping their overweight or buy ratings on the stock. They argue that, if anyone can survive, grow and make money in this industry, it will be TSMC.

Five years ago, the opposite appeared likely to happen.

TSMC and United Microelectronics (UMC), its smaller Taiwanese peer, invented and built up the business of contract chipmaking. The foundry industry, as it is known, thrived for more than a decade before the Taiwanese groups started facing competition. Chinese companies moved in at the low end of the foundry business while IBM started offering services at the top end.

By 2003 iSuppli, the market research group, warned that TSMC’s market share was slipping. Media reports started predicting that IBM would muscle its way to the top whereas TSMC and UMC would be squeezed out.

IBM had succeeded in snatching orders from some key TSMC and UMC customers such as Nvidia, Xilinx, Analog Devices, AMD and Broadcom. Meanwhile, Semiconductor Manufacturing International (Smic), China’s largest contract chipmaker, squeezed in to rank fifth among the global foundries.

Robert Tsao, then UMC chairman, even predicted the demise of the existing foundry model.

But, says Ming-kai Cheng, head of technology research at CLSA, “they were all wrong”.

China has not become a meaningful player in chip manufacturing nor have big integrated players such as IBM captured leadership in the industry.

Instead, TSMC has now pulled far ahead of everybody else. Over the past five years it has been the only foundry that has managed to hold on to its market share, remain cost-competitive, enhance technological capabilities and shield its profits from cyclical downturns.

Due to TSMC continue expanding their service and focusing on developing their capabilities, Mr. Cheng was quoted as saying "TSMC now looks more like Intel than like UMC.." The FT also says IBM have lost 3.5% of their market share (6.1% to 2.7%) and that both Chartered, the Singapore foundry, and UMC, the no. 2 foundry are both far behind TSMC. The FT further argues the low cost Chinese foundries are beating each other out of the market and that soon there will be fewer players. The FT also looked at the deregulation of investments in fabs in China (seeMr. Ma please don't move the fabs to China) saying it won't make much of a difference since UMC has apparently said China bound investments are not a priority. Afterall, as we argued in No 10G AUO Plant in China , China-bound investments are still strategic business decisions, not legaslative.

TSMC without doubt single handedly changed the landscape of the semi-conductor industry. They are one of Taiwan's best companies and continue to lead in the semi-conductor fabrication industry. Taiwan should be proud of this world class company and I certainly hope sensibility prevails and that both TSMC and UMC will not charge into China as soon as the new laws are passed. Histroy suggests TSMC have made the right moves at each stage of the game in their 20-year history. Lets hope they continue to do so.

Financial Times: TSMC set to forge a path through gloom

TSMC sees good Q2 but Expects Slowdown.

There is a lot of news about TSMC out there today. First up is there Q2 earnings report. According to a press release:

TSMC today announced consolidated revenue of NT$88.14 billion, net income of NT$28.77 billion, and diluted earnings per share of NT$1.12 (US$0.18 per ADS unit)for the second quarter ended June 30, 2008.

Year-over-year, second quarter revenue increased 17.6% while net income and diluted EPS increased 12.9% and 16.3%, respectively. On a sequential basis, second quarter results represent a 0.8% increase in revenue, an increase of 2.2% in net income, and an increase of 2.1% in diluted EPS. All figures were prepared in accordance with R.O.C. GAAP on a consolidated basis.

Second quarter business saw an improvement from the previous quarter, although revenue and margins continued to be negatively affected by the strength of the NT dollar. Despite the negative impact from the foreign exchange rate, our margins have exceeded guidance due to significant cost improvement and higher levels of wafer movements. Second quarter gross margin was 45.6%, operating margin was 34.5%, and net margin was 32.6%.

Despite posting good results for Q2, TSMC are expecting declining macro-economic factors to impact on their performance over the next quarter. CFO Lauro Ho said:

Even with a weakening macroeconomic environment, our second quarter results were either in line with or slightly higher than guidance announced at the end of April, thanks to our continual effort in driving cost reduction and increasing utilization rates," said Lora Ho, VP and Chief Financial Officer of TSMC. "Based on our current business outlook, management expects third quarter revenue growth to be below seasonality.

Reuters notes TSMC sees only sees a growth of 4% in the semiconductor industry. According to Reuters:

Top contract chip maker TSMC (2330.TW: Quote, Profile, Research) said on Thursday it expects the global semiconductor market to grow 4 percent in 2008, compared with a previous forecast of 4 percent to 6 percent.

According to PC World any slowdown in the foundry industry will have a widespread economic impact:

The warning from TSMC raises concerns about the global IT industry. The company is considered an bellwether for the global technology industry because it produces chips for such a wide range of products, including digital cameras, music players, mobile phones and PCs. Since chips are the building blocks of all electronics products, a slowdown in the sector will reverberate across the entire industry.

As a result of the slowdown, many fabs are cutting down on their CAPEX. Electronics Weekly notes even though TSMC is expanding, this expansion may not be enough to meet global demand and therefore expects potential supply issues in 2009. According to Electronics Weekly:

TSMC is still adding capacity, despite widespread industry concerns that the foundry industry is cutting back on capex in order to squeeze higher prices per wafer.

According to TSMC's Q208 report, the company currently has the capacity to run 2.3m 8 inch equivalent wafers which is 6 per cent more than it had in Q1.

TSMC's intention is to raise the Q2 figure by 5 per cent in Q3 to reach capacity of 2.4m 8 inch equivalent wafers

Total capacity for 2008 is expected to reach 9.4m 8 inch equivalent wafers, which is 13 per cent up on the 8.3m wafers processed in 2007.

12 inch wafer capacity in 2008 will increase by 27 per cent in 2008 over 2007, says TSMC.

Despite this, according to Europe's leading semiconductor analysts, Future Horizons, capacity won't meet demand next year.

"It's the first time the semiconductor industry has cut back on capital investment at a time when capacity is at its tightest", says Malcolm Penn, CEO of Future Horizons, "they're saying: 'I've got too little capacity so I'm stopping investing'. While utilisation rates are high, the capex rate is low."

However, to remain competitive and to allow companies to transition easily from 45 nm processing technologies to 32 nm processing technologies, TSMC has rolled out a new design platform. According to EE Times India:

Feeling the heat at the 32nm node, Taiwan Semiconductor Manufacturing Co. Ltd has put the pedal to the metal and rolled out a new design-for-manufacturing (DFM) scheme. TSMC's Unified DFM (UDFM) architecture is touted as giving chipmakers unprecedented access—at no charge—to its proprietary simulator, DFM models and other production information for 32nm. The goal of is to give them a head start with TSMC's 32nm process, which is expected to move into production by the end of 2009.

Other leading-edge foundries—including IBM Corp.'s "fab club" (AMD, Chartered, Freescale, IBM, Infineon, Samsung, STMicroelectronics and Toshiba) and United Microelectronics Corp.—are taking similar steps on the design automation and DFM fronts, hoping to make the 32nm transition somewhat less painful and costly.

As I said, a lot of news about TSMC out there today. I think that covers most of it!

Reuters: TSMC sees global semicon market growing 4 pct in '08
PRN News Wire: TSMC Reports Second Quarter EPS of NT$1.12
PC World: TSMC Q2 Results Shine, but Says Business Is Slowing
Electronics Weekly: TSMC Adds Capacity but is it Enough
EE Times India: TSMC rolls out new DFM scheme

40nm GPUs Coming Soon

TGDaily reports Taiwan Semiconductor Manufacturing Corp. (TSMC)will soon take the lead in chipset development when they roll out production processes for 40 nm graphics processor units (GPU). Although the development of 40 nm processing technology is fairly old news (we blogged about it in 40nm Processing Technology at TSMC), TGDaily offers a fairly interesting perspective. TGDaily writes:

Intel is proud of its dominant position in semiconductor in production technology and especially the fact that, for as long as we can remember, has led the industry in terms of the smallest chip structures. Its 45 nm technology is still at least one year ahead of AMD. But it appears that Intel will have to give up that lead next year, at least for a few months, when GPUs will make their transition to 40 nm.

This was bound to happen sooner or later. After speaking with several of our sources at ATI (AMD GPG) and Nvidia, we were told that a 40 nm GPU manufacturing process is on the way for first half of 2009. In fact, both companies are working on parts that should capture the spotlights at CeBit 2009 in Hannover. Both low-end and mainstream products are ready to be manufactured in 40 nm soon and should be on display at the tradeshow.

It appears that TSMC’s previously announced $10 billion investment in manufacturing technology is yielding results already, since the company is now able to develop 45 nm and 40 nm processes at the same time. The next step for TSMC is either 32 nm or 30 nm - or below. Samsung is investing heavily in 30 nm, but that is for DRAM only.

Intel has 32 nm CPUs still in development at its research, development and production facilities in Hillsboro, Oregon. 45 nm Nehalem CPUs will be the focus at the upcoming fall IDF, but it is generally expected that prototype 32 nm processors will be first shown at the company’s spring developer forum in H1 2009. Production of the chips should begin early in H2 2009, with volume shipments beginning in late Q3 or early Q4. First chips should surface in commercial products in late 2009, while 32 nm will be a 2010 topic for the mainstream buyer.

However, by that time, millions of 40 nm GPUs will have shipped already and you can bet the farm on the fact that both AMD and Nvidia will be pitching that story to the media in the same way Intel did in previous years.

Its good to see a Taiwanese company taking the lead for a change. Although Intel will catch up soon, it shows Taiwanese companies are as innovative and competitive as ever. TGDaily sums it up best:

Kudos to TSMC for developing the 45 nm and 40 nm half-node die-shrink at the same time.

TGDaily: Intel to lose its lead in chip manufacturing tech in 2009, sort of