GLG News by Israel Beinglass, Ph.D.

Implications: With the skyrocketing cost for developing new processes more and more semiconductor companies are joining the IBM alliance the pre competitive process is getting less and less. NEC is just the last one from a parade of the companies that already joined the alliance. Looks like we are having only 2 classes of semiconductor technology in the world right now: a.Intel  b.IBM alliance. The rest of the world semiconductor manufacturing is either part of it or copying from it. The most benefited company from that arrangement is IBM since all the payments received from the alliance members goes practically to the bottom line.

Analysis:

The rising cost of developing new semiconductor process technology node (65nm, 45nm, 32nm etc) has been risen substantially in last decade. It cost over $500M and 1 to 2 years every time a company is developing new process technology. Also even after spending the necessary amount of cash on the new process it is not guarantee that the process will be competitive enough with the rest of the industry. It can easily fall  short of the needed performance compare to the competitors.
IBM developed the alliance model where by companies joining (for a hefty fee) the alliance they can send a group of engineers to IBM facility as a part of the development team with IBM engineers and the rest of the alliance participants.
The following companies are member of the alliance:AMD, Chartered, Samsung, ST, Toshiba, freescale, Infineon and NEC the latest to join.

Looking at the global map of semiconductor logic development, world wide, looks like we have only two super power:Intel and IBM alliance, the.
 The trend is a great one for IBM (Note IBM stock price dropped very small amount compare to the rest of the market) since most of the money they are getting from the alliance members goes to the bottom line. On the other hand it helps the member companies to control their R&D cost and spending.
 The alliance saved Chartered that was almost on a brink of bankruptcy and brought it back to the main stream foundry.
Another big loser is the semiconductor equipment manufacturers. Instead  of developing technology in 9 different site (a lot of tools sales) the development is been done in one site only (IBM). Also for the equipment manufacturers it is very important to win the IBM development status since the alliance members will be using the same equipment when they transferring the process to production in their own facility.

Implications: The DRAM market is in a middle of a major down turn, over capacity and price decline plaguing the industry for the last year. Samsung exec for investor relations claims that toward second part of next year we will see price firming. The question is it really going to happen within that time frame.

Analysis:

The DRAM market is a commodity market and like all other commodities it is cyclical and depends on among other things on global economy. The demand for DRAM has constantly going up as new products that need DRAM are introduce to the market. So far so good, however the market could be flooded by over supply (Fab capacity) since most of the DRAM suppliers are trying to grab market share when prices are high and they are falling into the trap with their Fabs running when prices are falling.
During the last 3 years there were many 300mm Fabs that started making DRAM as well as more companies are trying to get more market share (Samsung, Hynix, Elpida, PoweChip, Promos, Quimonda, Micron) process that ending in huge price erosion.
Luckily one new parameter that affected this cycle is the growing demand for NAND Flash that offset some of the problems with DRAM pricing. Some of the companies (Samsung, Hynix) were shifting production back and forth between DRAM and NAND.
However even that didn't help and the traditional cycle is now in a full swing, while at the same time new production is still coming on line.
Samsung claims that the down turn will be over late 2008, eventually it will happened, though I think more likely by early 2009.

One wild card that is important to consider is the world economy that might getting into recession...in that case all bets are off and it might take one more year after 2008...

Implications: The article point a very important issue, there is no hope for Micron to compete in the DRAM market with the Asian companies (Samsung, Hynix, Elpida). As all the other US companies (sadly) got out of the memory business, Micron will have to exit this market.

Analysis: For years Micron stand as the lone ranger trying to compete in the DRAM market with very limited success. The DRAM market is very difficult market to compete and Micron gave a very admired fight, but it's time to move on.
Most of Micron's fabs are aging 200mm, they are totally not competitive with the new 300mm Fabs in Asia. The fabs will have to stop making DRAM very soon. At this time the only alternative for Micron is to continue with the COS (CMOS Image Sensor), though even there they are facing tough competition from OmniVision and the other competitors in Asia that some of them are turning to manufacturing in 300mm Fabs.
The IM (Intel Micron) JV is too little and too late, the NAND market is totally dominate by Samsung, Hynix and Toshiba/Sanidsk, so IM has very little chance to be a major player.
Overall I don't think Micron will survive as an independent company for more than few years...

Implications: Samsung has been loosing market share during the last few quarters, by increasing capacity they would like to fight back especially Hynix and Elpida and regain market share....however the article mentions that it could be a devastating process since the DRAM market is already in a big decline in ASP, so any extra capacity will accelerate this issue.

Analysis:  The DRAM pricing continue to drop due to over supply and very aggressive price cutting from Samsung's competitors. Looks like it did work for a while since Samsung who became somewhat complacent for the last year lost some market share. It is true that the situation in NAND is somewhat different since the demand for the product is still very high, but all Samsung's competitors are adding more and more capacity for that growing market.
Samsung's decision to hike their CapEx spending in order to regain market share is good news to the different tool makers (AMAT, NVLS, LRCX, ASML etc) but could act as a boomerang and get the price in a tail spin to lower and lower, at this time everyone loses....

Implications: The article trying to show on a positive note that they are forming a consortium to encourage other semiconductor manufacturing to increase the adopting rate of SOI technology. However the fact is that the usage is not growing and SOI is still a niche application and will stay like that for a while.

Analysis: SOI technology has been around for over 30 years as a great idea that has been always an idea for the future. Late '90 IBM came with new approach and decided to bring the SOI (Silicon on Insulator) technology to the main stream of semiconductor manufacturing.
The move was a bold move that was taken by surprise with most semiconductor manufacturing. The use of SOI wafers (the raw material that the chip is built on) and the technology have been very slow to be adapted by other manufacturers.
As of today only IBM and AMD are using it in a large-scale manufacturing.
The reason why the consortium was formed is to bring back the excitement to the technical community in order to increase the awareness of the technology.
One of the key issues of using SOI wafers is the cost of the raw wafers themselves compares to the standard Silicon wafers. Also when a device is built on SOI wafers there are some major changes that needed to be done on the design side as well as on the manufacturing process.
But the bottom line is that the end results of using SOI wafers don’t improve the device performance substantially to justify all the extra cost and process and design changes.

Implications: MagnaChip has been under performing for several quarters, out of their three business sectors 2 of them (CIS and LCD Drivers) are dismal and only the third one -special foundry is doing sort of OK. The article mention another addition capability to their foundry business by cooperating with AMI.

Analysis: Magnox is trying to play in three different businesses, CMOS Image Sensor (CIS), LCD Driver and special foundry. The company's revenue has been going down for the last few quarters due to the price competition in the first two businesses.
In my mind the only part of their activities that has a chance to flourish is the special foundry. The market share that Magnox carved out is very impressive and since there isn't too much competition in it, there is a chance to be very successful.
The agreement with AMI gives magnox a boost in their foundry business that eventually if played right will bring some of the lost business back. This could be just the first sign of a turnaround...

Implications: The recent price drops in the DRAM market (and NAND) is an alarming issue.
Nobody knows where it will stop. Recent drops in Micron profitability and Hynix rising as number 2 vendor, could be a sign of price war where everything goes...and nobody wins.

Analysis:  

The recent price drops in the DRAM market (and NAND) is an alarming issue since the capacity built around the world of manufacturing is enormous and there is a reasonable fear among several analyst that the glut in capacity will cause two things: 1. continuation of price erosion and 2. reduction in orders of SemiCap . Both of these issues are getting very close to reality.
Recent built up of memory capacity is insane, about 70% of new orders of SemiCap are to the memory companies while the logic companies are holding back. Main reason is due to the conversion to foundries from many IDM's (foundries are more efficient) and due to the conversion to 300mm wafer size (there are about x2.5 more chips on 300mm wafer than on 200mm wafers).
Looks like the memory companies haven't learnt from previous down turn and price sliding.

Implications: The article discusses the increase demand of Apple to NAND memory chips. Based on the increase demand for iPod and the upcoming iPhone Apple needs all the NAND chips it can get his hand on, Samsung and Hynix will be the main beneficiaries of that.

Analysis: We are all aware of the glut in the memory market that we have been witnessing during the last few months. Both DRAM and NAND pricing were plummeting more than 60% and the concern was and still is that the trend will continue and eventually cause the memory market to collapse.

We are getting first glimpse of hope for continuing strong memory market from Apple. Based on the article, Apple will continue the brisk demand for NAND memory chips in order to fuel the demand for iPod and the upcoming demand for the most anticipated iPhone.
If this is true than the NAND demand will continue to grow, bringing some stabilization to the bitten up market.

Implications: Capacity in the foundries is going down for the last quarter.
We need to watch this parameter very careful since it might be the first sign of a real slow down in semi activity.
If the trend will continue for next quarter or 2 quarters things might get ugly for the industry.

Analysis:

The huge capacity expansion in the Semi industry is coming (maybe) to a halt. The expansion (especially in the memory segment) that we were witnessing during the last two years is ending. As most analysts say the second part of 2007 will not be good for the industry (and to the Semi Cap Equipment as well).

It is quite interesting the view of the article that discusses only the foundries and the slow down that they experiencing. The foundries did not expand their capacity as much as the memory manufactures did, albeit their utilization has degraded. This could be an early sign of more severe problem that will haunt the industry for next few quarters.

There are really only three worth mentioning foundries in the world (TSMC, UMC, Chartered). I find it quite amazing that the whole world supply of foundries capacity could be filled by these three companies, especially when even they can't fill their Fabs capacity.

The bottom line is that even though the memory segment of the market is continuing expanding like there is no tomorrow the logic side as reflected by the foundries business starting to shrink. The day for reduction of capacity in the memory business probably just around the corner.

Implications: "Old Fabs never die"... they always find new buyers. Recently few "used" Fabs were exchanging hands, selling of ZMD Fab to X-Fab is the latest in the trend.
Looks like for old Fab there is always someone who would like to buy it for the right price...

Analysis: Recently several "old" Fabs were exchanging hands, noticeable the sell of LSI logic 8'' Fab to OnWafer, and the selling of the 6" Fab of ZMD to X-Fab (in the above article).
A 6'' Fab is typically very old-circa mid to late '80, it has become fully depreciated for a while with very low book value. In that case for the seller it is a great deal since upon selling the asset he recognized a real value, while for the buyer it represent a great opportunity to walk in the Fab and start running production paying low price.
The assumption is of course that the seller wants to dispose of the asset and the buyer is looking for an extra capacity.
All the 6" Fabs are running old technology nodes of probably >0.3micron. In that case the buyer is a producer of analog or high voltage chips where the need for advanced technology node is not required.
The price of that deal has not published but base on my experience and other deals on 6" Fabs the price would not be more than $30M.

In other development regarding X-Fab looks like they are getting more into the analog foundry business in a big way competing with
MagnChip (Korea) on the analog chip foundry business.

Implications: The dropping of the NAND flash is a bad news to the semiconductor manufactures but is good news to the system manufactures, since they can stuff more memory for less.

Analysis: The additional of Fab capacity during the last few years especially in the NAND memory was astonishing. Samsung, Toshiba/Sandisk Hynix and Micron were adding fab capacity like there is no tomorrow.

The semiconductor Cap Equipment like Applied Materials, Novellus and Lam benefited greatly from this built up. However as it is now this will boomerang back since the expansion will stop till the world access capacity will have to work its way out.

The system manufactures like hand held devices and cell phone also benefit from the NAND price dropping since they can add more memory storage to their devices for less money, as the article put it the price is almost free...

The trend toward hybrid PC where some of the operating system will reside in Flash memory will accelerate with the price drop. Soon we will able to have an instantaneous booting of our PC vs. long booting with current hard drive technology.

The ultimate price of the NAND is quickly approaching $1/Gig...

Implications: In 1982, Applied bought a UK Implant company (Linttot) that became the Ion Implant division.
The division was struggling for many years with only few golden years (1998-2002). Recently, the division was losing market share and became a burden on Applied P&L leads to management decision to close the division.

Analysis: The Implant business has three different segments:
1. Medium current
2. High dose-low energy
3. High energy

The high energy is relatively small segment and is controlled by Axcelis. The medium current is controlled by Varian. The high dose-low energy is the most challenging sector with the highest growth rate.

Applied was one of the market share leader in the late 90's early '00 with great batch tool that was sold everywhere. The market condition was demanding the transition to single wafer tool in order to have a better process control. Applied answer to this didn't satisfy the market and they were losing most of the market share.

Varian, in the meantime, was perfecting their single wafer tool and became the market leader within the last few years. On top of that, running a division in the UK is very costly and with a shrinking market share there was no other choice but to close the operation. Varian and to a lesser degree Axcelis will the 2 companies that will be gaining market share.

In a way, Varian becomes almost the only game in town...

Implications: The article put it rightly so that the CapEx will drop in 2007 - the semiconductor industry had gone through a huge expansion during the last two years, especially in the memory side of the business. Slow down in the chip technology scaling and extra capacity will slow the CapEx in 2007.

Analysis: During the last 2-3 years, the semiconductor industry had gone through a major expansion in new 300mm Fab facilities.

Several new Fabs are ready to start production in 2007 mainly in the NAND Flash (see Toshiba/Sandisk, Samsung, Hynix and Intel/Micron) and DRAM.

The huge memory lines that will be ramping production through the end of 2007 will saturate the world demand and will create pressure on chip prices that will lower companies profit.
Lower profit of the semiconductor companies will have an adverse effect on demand for new equipment hence the CapEx will be lower.

Also, a lot of the new devices that are used in hand held devices have relatively small area that doesn't require a lot of Silicon area that in turn wouldn't  maximize any demand for big increase in silicon capacity.

Implications: Applied Materials has gone through a massive stock repurchase during last few moths using a lot of the available cash in the program. Question is why and what will it do to the company long term.

Analysis: Applied Materials was buying back their stock at a very high rate during the last few months. Several issues could be address regarding this
1. Applied Materials had a lot of cash that was generated during the last decade (over $6B), the huge amount of cash became a liability for the corporation since it could be a target for a hostile take over and the fact that there was no use for it and it just sitting and getting money market rate in the bank.
2. There was no major acquisition that the company gone through that needed a lot of cash.
3. The company as the largest company in the business was not able during the last few years to get ahead (they are good in what they are and no improvement in their weak position). Based on that wall street wasn't that enthusiastic about the stock hence stock price was lingering compare to some of the other players (see VSEA, LRCX).
4. An easy solution was to use the pile of cash to repurchase stocks in a relatively low price and enhance share holders value. In the long term when hopefully the stock price will rise it would be recognized as a smart move, while they still maintain a healthy amount of cash.

Implications: The CMP (Chemical Mechanical Polishing) business generates an unbelievable consumable business that is as big on yearly revenue to the size of the hardware sale ~$1.1B. This is an interesting situation and very lucrative very high margin business that very few companies are participating in

Analysis: During the last few years a whole new breed of companies that serving the consumable market appears in the landscape of the semi CapEx. These are the suppliers of the consumable for the CMP machines that control by Applied materials and Ebara. Two kind of consumable, Pad and Slurry are enjoying huge market of over $1B/year with Rohm&Hass (Pad) and Cabot Micro (Slurry) with nice yearly growth.
With the industry moving to the Copper technology in the logic arena the Copper slurry became very important component in every advanced fab with great pricing. However the next step in the Copper slurry just started when the huge memory market (DRAM and FLASH) are in a process of converting from Aluminum technology to Copper technology. This change will make the change from Oxide CMP with low margins to the more advanced Copper CMP with higher margins.
Due to that the companies supplying the slurry will be doing well in the years to come.

Implications: The semiconductor industry is going through many changes, it is maturing, also the relentless race to faster and faster devices leaves the designer with ever changing process that is looking for next application

Analysis: Who didn't hear the "what's the next big thing", the reason for that is that the semiconductor industry is on a constant race towards new and improved devices. This was propelled by three items.
1. faster and faster Microprocessor
2. denser and denser memory
3. PC with better and better Operating system.
However with Moor's law getting to the end and the industry maturing many changes are happening in the industry, one of them is the private equity taking charge and grabbing more than ever was taken by them. One of the reason is the VC community can not provide the huge return of 10-20x as in the past.
The bottom line embrace yourself to more mature industry with only several "high fliers" and more mundane returns.