Description

ChipMOS TECHNOLOGIES (Bermuda) LTD. (IMOS)

ChipMOS is an outsourced testing and assembly service provider for semiconductor manufacturers that I believe has minimal downside due to a ridiculously low multiple on a growing business and upside of 100-200% over the next year with a conservative multiple. The stock is currently trading at $5.70, and will earn $0.58 this year. This year is a depressed year for the business and the company has recently added significant new capacity which is currently underutilized. The market is trading the stock at a low multiple of depressed numbers despite the fact that the last several months have seen definite signs that the business is improving. Making conservative assumptions, I project earnings of $0.87 in 2006 (consensus $0.78), with potential for further significant improvement in subsequent years. So, it's cheap in absolute terms: 6.5x p/e based on next year's (growing) earnings with a moderate debt-load. Further, comps are trading at 15-25x 2005 earnings and well over 11x next year’s earnings - implying at least $10/share to ChipMOS on a relative basis.

But the story could get better because ChipMOS depreciates more aggressively than its comps creating approximately 50cents/share of pickup in our estimation and the company has a competitively positioned plant in China.

The weak 2005 has masked the growth and growth potential of the company. I believe that in 2006 both could manifest themselves very clearly. The company is misperceived to be highly cyclical because of their exposure to DRAM. While it is actually relatively stable because of their long-term contracts and volume-driven nature of their business model, as opposed to classic semiconductor manufacturing boom/bust cycles of pricing and capacity.

Taiwan based ChipMOS TECHNOLOGIES (“ChipMOS”, “IMOS”, or the “Company”) is the world’s 5th largest outsourcing provider of testing and assembly services for semiconductors and is the largest within its niche focus in memory and LCD Driver chips. The stock trades at 6.5x 2005 (calendar=fiscal year end, so mostly in the bag) EBITDA less Maintenance CAPX (“E-MCX”) and 2005 is a depressed year for the business due to a confluence of factors. I believe the stock is trading at 4.3x a conservative estimate of 2006 E-MCX. The Company has a favorable 15% tax rate which makes these numbers comparable to 4.6x 2005 E-MCX and 3.0x 2006 E-MCX for a full US taxpayer. The Company’s recently completed a new manufacturing facility in China which is underutilized and will allow for years of growth and at reasonable returns on capital (15% after-tax). At 8x 2006 E-MCX (which based on the 15% tax rate implies 10x cash earnings and 5.0x EBITDA) the stock would trade at $13. As the business turns (and I believe it is already showing signs of being in the first stage) and the market gives it a moderate growth multiple of 11x E-MCX (14.6x cash earnings), the stock would trade at $18 along with generating a $1 of cash implying an $19 value from today’s $5.70 or a 200%+ gain with little downside risk.

Near term catalysts could prove my estimate of 2006 EBITDA to be conservative. Valuations of other “back-end” providers suggest that the market is willing to pay substantially higher multiples for weaker companies.

I also believe, though obviously don’t have facts to back it up, that due to the company re-striking the price on their convertible notes (this was a one-time event and cannot happen again) convertible funds have been shorting the stock over the last few weeks to adjust for their additional shares due to a new lower strike price. This could be a reason the stock has not reacted to the very positive Spansion announcement and continued strong monthly revenue numbers. Further, I believe the market has also misperceived the recently filed shelf offering as a sign that the company is in someway capital constrained and needs to go back to the market for additional financing.

BUSINESS DESCRIPTION

ChipMOS provides testing and assembly services for memory (approximately 70% of sales) and LCD Driver (20% of sales) (“LCDD”) manufacturers. The remaining 10% of sales is mostly testing/assembly of mixed-signal/logic chips. IMOS is the leader in memory and LCDD testing/assembly and largely does not compete with the “Big 4” Outsourced Assembly and Testing Service providers (“OSATS”), such as AMKR, STTS, ASX and SPIL. It is not likely that the Big 4 would enter IMOS’s market because their equipment is different and therefore they would require huge capital expenditures to have the scale necessary to compete with ChipMOS.

Simplistic definition of testing and assembly:
- Testing: making sure the chip does what it is supposed to do. Both functional and speed tests are performed. Wafers are tested (wafer probing) and final chips are also tested.
- Assembly: All the steps from taking a wafer, that can hold hundreds of individual dies, and making each one into a chip. e.g. cutting the wafer, adding leads and molding, etc.

Testing is generally charged by the hour, and assembly is by the piece. Customers will have specific piece of testing equipment that their chip was designed to be tested on and typically will have their own proprietary algorithm to test their chips.

ChipMOS’s niche focus has allowed them to enjoy scale economies evident by their relatively high margins. ChipMOS has about 40% of the total DRAM back-end capacity in Taiwan and about 45% of the total LCDD testing/assembly capacity in Taiwan. Also contributing to their superior margins are their long-term contracts (discussed herein) which reduces SG&A.

IMOS is a Bermuda holding company which essentially owns 70% of ChipMOS Taiwan and 100% of ChipMOS Shanghai. The corporate structure is more complicated than this - one should check the 20-F for more detail. Currently most of the revenue is coming through Taiwan; however most of the growth will come from China.

ChipMOS Taiwan was founded in 1997 as a joint-venture between Mosel Vitelic and Siliconware Precision Industries Co. (SPIL). At this time ChipMOS was the back-end operation of ProMOS, and was an affiliate of Mosel. When ChipMOS Bermuda was formed in 2000, Mosel exchanged their shares in ChipMOS Taiwan for ChipMOS Bermuda; however, SPIL chose to maintain their ownership in ChipMOS Taiwan. Currently Mosel is a holding company that holds 40% of the ChipMOS Bermuda shares and approximately 10% of ProMOS (our largest customer). ProMOS is the second largest Taiwanese DRAM manufacturer. In short, the public entity which I am discussing here is IMOS or ChipMOS TECHNOLOGIES Bermuda whose float is 40% owned by Mosel. The public entity owns 70% of ChipMOS Taiwan (with SPIL owning the remainder) and 100% of ChipMOS China.

The company trades on the NASDAQ and provides US style disclosure in the forms of 20-F's and 6-K’s. I understand people may put a discount to multiples based on that fact, but we would note that in past years when the business is strong that discount quickly disappears, so we believe this may just be a down-cycle factor. Of the comparables, STTS, SPIL, ASX, and ASTSF are all similar in their disclosures

The Company has invested over $150 million (half of which for the physical PPE) in a 750,000 square foot facility in Shanghai, with a 180,000 sq foot Class 1,000 Clean Room (1 dust particle per 1,000), which has just begun to produce revenue. China is important because it is where the customers are going. It also should be noted that because of the difficulty Taiwanese companies investing in China most of the other Taiwanese testing and assembly providers do not have any capacity in China. IMOS employs over 5,000 people.

Memory is used in computers, cell phones, mp3 players, etc.
ChipMOS is largely agnostic to the type of memory that is being tested. Their equipment can be used for NOR and NAND flash as well as all the flavors of DRAM.

LCDD’s are chips that control LCD panels. They are a critical piece, but only account for a small portion of the panel cost.

Long-term contracts.
Over 50% of IMOS’s volume is in some form of a take-or-pay contract. Most of these contracts provide volume minimums at “market prices” subject to a floor equal to IMOS’s depreciation and operating costs. ChipMOS’s customers include the two largest DRAM manufactures in Taiwan (ProMOS and Powerchip), the Largest LCDD manufacturer in Taiwan, Novatek and now also includes Spansion. These large contracts contributes to IMOS’s economies of scale, by being better able to keep utilization rates high, and by lower SG&A expenses.

Outsourcing.

The trend over the last several years has been to outsource backend services. Most industry analysts expect this trend to continue. In addition to the Fabs-less manufacturers, who obviously have to outsource, many foundries have fabs, but no back-end (“back-end”-less).

This trend is not likely to reverse for a number of reasons: (i) the relative high costs in making wafers makes it difficult for manufactures to have scale in the front-end as well as the back-end, (ii) although the back-end is extremely precise and requires sophisticated equipment in very expensive clean rooms, relatively speaking it is reasonable to understand why a DRAM manufacture might consider the back-end part of the process to be more commodity oriented and therefore want to outsource it, and (iii) without sufficient scale, manufactures would not be able to perform their own back-end at any cheaper prices than they can outsource it for. Further, the adoption of DDR2 (described below) should accelerate the outsourcing trend, because of high cost of the DDR2 equipment.

Taxes
The company receives certain credits for continued investment and R&D in both Taiwan and China. According the company, they expect their long-term tax rate to be between 10 and 15%. Although in 2006 it should be in the mid single digits because of utilizing their NOL and other finite credits.


NUMBERS

83 shares (a)
$5.70 price per share (Thursday’s close)
$473 Market Value
$226 Net Debt (b)
$699 Enterprise Value
(a) Fully diluted and assumes conversion of convertible bonds.
(b) Includes cash, investments and minority Interest, and PV of NOL (approx $30mm).


Operating Results / Projections
2002 2003 2004(c)2005e 2006e
Revenue: $141 $225 $435 $466 $558
EBITDA: 55 102 195 212 262
DEPR.: 81 81 106 145 150
E-MCX(d): - - 95 112 162
CAPX: 69 74 245 127 164


(c) To normalize the effect of the NTD/USD exchange rate, all above operating results are translated by using an exchange rate of 33.475 NTD/US$. 2004 also excludes ChipMOS’s low margin Turnkey business, which they mostly phased out earlier this year.
(d) EBITDA less maintenance CAPX of $100mm. Maintenance CAPX could actually be as low as $60mm.

Looking at the above numbers, please note that 2005 is a down year for the industry and IMOS grew in spite of it because of the growth CAPX. 2006 starts to benefit from use of China.

DEPRECIATION AND MAINTENANCE CAPX

Maintenance CAPX (defined as the amount of CAPX required to keep revenues constant) is a difficult number to measure because of the fluctuating nature of testing ASPs. The assumption I have made is based on the economic life of their testers being 7 years (may be as much as 10), which they generally depreciate over 5yrs. The Company uses their fully-loaded costs (including depreciation) as a firm data point in negotiating ASPs. To this end, it is in the Company’s interest to adopt more aggressive depreciation schedule. As a point of reference, depreciation policies of the other OSAT’s are: STTS(8yrs), AMKR(4-6yrs), SPIL(1-9yrs), ASX(3-8yrs). Trying to back into the competitor’s depreciation schedule from public filings, yields an average 7 years.


I believe that $100mm is a conservative estimate of this number. The Company believes their maintenance CAPX (defined as I do), is $60mm per year. Using similar Assumptions I estimate ROIC to be at about 17%. 2006 CAPX should be $130-$150 million due to continued capacity expansion plus – this amount excludes the capacity that will be needed to satisfy Spansion.


2005, what Happened - Why are we flat on the year vs. 2004?

There are a number things that have gone the wrong way for IMOS in 2005:

1. ASPs for DRAM testing have been lower than expected. 2005 DRAM ASPs are approximately 10% lower than 2004, i.e. on an apples-to-apples basis, 2005 Revenue would have been at least $50mm higher than 2004. I believe that ASPs will only rise when it’s clear we’ve reached the bottom for DRAM pricing. DRAM manufactures are now selling product at or near their cash costs.

2. Approximately $25mm in CAPX was spent on 5 Advantest T5593 high speed DDR2 testers that currently are being used to test slower DDR1 chips and high speed graphics chips, which is akin to using the color copier to make b&w copies. There have been a series of events that have delayed the widespread adoption of DDR2. I have spoken to many people in the industry and it is a question of when, not if. I currently believe that 2006 will show improved DDR2 volumes. To the extent that the transition continues to move slower than expected, my price target would likely remain unchanged, however my timeline would be extended.

As DDR2 becomes mainstream, the following would happen:
a. ChipMOS could charge higher hourly rates for its high speed testing equipment
b. Average testing times would increase. Initially, 2,000 seconds vs. 200-300 seconds for DDR. And hence increase utilization rates.
c. Increased packaging revenue. DDR2 costs $0.50/chip vs. $0.25 for DDR.
d. Increased outsourcing from Integrated Device Manufactures (“IDM”). Many IDM’s that do their own testing do not have sufficient high speed testing capacity in house.

3. Exchange rate – The Company’s business is run in NT Dollars, however the shares trade in USD on Nasdaq. The NTD has depreciated by 6% since the beginning of the year and 9% since March, and therefore has made the US$ adjusted earnings less. Given their guidance is in USD, the depreciated NTD made it impossible for them to hit their 2005 numbers, which seemed to upset the analysts.

4. Utilization rates below expectations. Memory testing utilization rates were approximately 80% for YTD 2005, vs. 89%, and 88% for full year 2004 and 2003, respectively.


5. Management Distractions. Management was very busy in 2005. Here are some of the things that they were doing: Merged two subsidiaries, opened Shanghai Plant, dealt an independent investigation surrounding 2 former directors and crimes they were accused of in the early 90’s at another company (they are no longer with IMOS), and negotiated and signed major contract with Spansion for NOR FLASH.

Cyclicality?
ChipMOS’s customers are DRAM manufactures, which are in a very cyclical industry. However, IMOS’s exposure to this cyclicality is mitigated. While there is definitely pressure from customers to lower prices when they are struggling, IMOS’s business is mostly volume driven. DRAM manufactures will cut testing costs, by reducing testing times (less stringent testing), but every chip is still tested and every chip is still assembly. To that end, when DRAM is in a down-cycle, we expect to see reduced testing times (evident through lower utilization rates. Approximately 79% YTD in 2005 vs. 89% in 2004 – significant for a high fixed cost business) and some pressure on testing ASPs, but will not see the dramatic peak-to-tough fluctuations that you would see from the DRAM manufactures.


2006 outlook

2006 should be a better year than 2005. The LCD driver side of the business is strong with rising ASPs, and on the memory side of the business: Flash is a growing part of their business (especially with Spansion) and the addition of more DDR2 business will add at least $50 million in revenue. ASPs for testing on the Memory side of the business were down 10-15% this year vs. 2004, which I believe was closer to an average year. Further, additional capacity that has come on-line will add to the top-line, even without the wide-spread adoption of DDR2.


CATALYSTS

There are a number of catalysts that should propel this stock forward in 2006.

1. DDR2 transition. I expect DDR2 to add at least $50 to the top line. Because IMOS is largely a volume driven business, technology transitions such as this are good for IMOS. New generations of Memory e.g. (DDR to DDR2), require longer testing times. Same is true when manufactures decrease die sizes (simply more chips per wafer).

2. Spansion. On Monday, the Company announced that they will provide testing and assembly services for Spansion, the Flash memory joint venture between Fujitsu and AMD which is expected to have their IPO this week. My estimate is that Spansion will initially add at least $60 million in revenue. This agreement is the 1st of what IMOS hopes to be 3 stages of agreements. While all the details haven’t been disclosed, this contract has the same take-or-pay characteristics of their other major contracts. Spansion adds another high profile name to IMOS’s customer list and is in a Memory segment where ChipMOS currently doesn’t do much business - NOR FLASH. (Note: Spansion volume is not incorporated in my 2006 estimates). Also note that as they diversify I believe people will stop labeling them as a DRAM tester and start to reward the company with a multiple, commensurate with the other OSATs.

3. Cleaning up corporate structure. The Company has recognized that their corporate structure is complicated and has taken steps to simplify it, e.g. merging two of their smaller subsidiaries into ChipMOS Taiwan. Additional steps could include moving SPIL from ChipMOS Taiwan into the Bermuda holding company, or buying them out.

4. China – Growth in Shanghai. ChipMOS is ahead of its competition in Shanghai. They are strategically located in the Qiomgpu science park, near many current and prospective customers. IMOS initiated their China project in 2000. They have spent $150 million so far in china. They space is substantially underutilized. The value of getting through the Chinese red tape and having ample room is tough to value but it clearly gives them a competitive advantage. I understand that generally speaking, facility engineering takes about a year, equipment setup takes about 6-12 months, and it takes about 3-6 months to qualify new customers.

5. Trend towards outsourcing. Outsourcing of back-end services in memory and LCD drivers is increasing. Testing equipment is very expensive and the back-end is a specialty in of itself. There are significant scale economies in this business.

6. DRAM price. It’s nearly impossible to predict the DRAM cycle, but any improvement in pricing would benefit ChipMOS. I believe that we are approaching the bottom as manufactures are selling chips near their cash costs.

7. Analysts. The Analysts that cover the stock (LEH, RBC, FBR, and TWP) have mostly taken down their numbers for 2006 to levels which I feel are extremely conservative. Upon speaking to them, it is clear that they are tired and frustrated with having to taken down their numbers (twice). To this end, they made 2006 projections they won’t have to be lowered. Raising estimates could prove to be another catalyst for the stock.



VALUATION
I feel that the company is extremely cheap on an absolute basis, and ridiculously cheap on a relative basis. While my investment thesis does not rely on comps, they do illustrate the tremendous disconnect between ChipMOS’s valuation and the rest of the industry (most of which have weaker operating results than IMOS).

LTM Debt / EV / EV / LTM P /
EBITDA Margin EBITDA Sales EBITDA 06 E
AMKR 246.6 12.9% 7.9x 1.6x 12.0x 25.5x
STTS 329.5 30.9% 1.8x 0.6x 1.9x 25.3x
SPIL 340.6 30.6% 0.7x 0.2x 0.7x 9.5x
ASTSF 164.4 32.2% 1.4x 1.8x 5.5x 15.2x
ASX 597.2 24.2% 1.8x 1.7x 6.9x 11.7x
---- ---- ----
Average 1.2x 5.4x 17.4x

IMOS 201.4 45.2% 1.1x 1.5x 3.4x 6.4x

To reiterate, as the business turns (and I believe it is already showing signs of being in the first stage, evident by the sequential growth in monthly NTD sales) and the market gives it a moderate growth multiple of 11x E-MCX (14x cash earnings), the stock would trade at $18 along with generating a $1 of cash implying an $19 value from today’s $5.70 or a 200%+ gain with little downside risk. Note, the discrepancy between E-MCX and p/e multiples is caused by the inherit catch-up between depreciation and capital expenditures, which is caused by the difference between economic and accounting depreciation. Additionally, my 2006 estimates could prove to be low and add further upside to these numbers.

It should also be noted that the market hasn’t always valued IMOS at these prices. During 2004 the stock traded above $15 which was approximately 10x trailing ev/ebitda. At first glance, the picture today looks like it did a year ago, i.e. trailing revenue of about $450-500, with growth expected in the NTM; however there are a couple of key differences: (i) 2005 is flat vs. 2004 despite falling ASPs and utilization rates, and today IMOS has more capacity (hence lower utilization rates), (iii) growth potential in China is here for 2006 and (iv) Spansion will add significant revenue and diversification. In other words, while the results looking strikingly similar between 2005 and 2004, they are not – 2005 is/was a depressed year.


While there is a fair amount of noise surrounding this story, I believe that underneath it all is a misunderstood, growing company and a bargain price. To be clear, if everything falls our way, I think we should get to $19/share, however, I am not so bold to bank my investment thesis on everything falling our way – if IMOS gets a reasonable multiple (which is also more or less in line with the comps) on my conservative 2006 estimate, the stock should trade above $10/share.

Catalyst

DDR2 Transition
Spansion Contract and revenue diversification
Cleaning up Corp. Structure
Continued expansion in China
Outsourcing trend
Improved DRAM pricing environment
Analysts raising estimates