Varian is three years into a five year paradigm shift from a member of a well defined and stable niche semiconductor equipment sub-sector triopoly to a holder of a near-monopoly. A critical technological shift and the consequences of the three companies’ strategic choices were visible 5 years ago, but were largely ignored due to the initial slow rate of change. However, this gradual shift rapidly accelerated in late 2005 culminating in early 2007 when AMAT was forced out and Axcelis became relegated to the vicious cycle of the AMD to Varian’s Intel. Nevertheless, since guiding down in October into dramatically deteriorating industry conditions, Varian is down nearly 50%, wiping out a year of impressive gains. This has created a fantastic entry point into a company quickly becoming a cash cow with a widening sustainable competitive advantage in a very attractive industry.

The market understands some of this but is underestimating its magnitude, sustainability and the corresponding valuation in part due to the focus on the current market weakness and near term industry outlook. As a result, Varian represents a compelling intersection of value, growth and sustainable value creation. While not a classic value stock, it is a deep GARP story where one can own a premium, industry dominating franchise at a fire sale prices. It's a Buffett stock if he invested in technology.

There’s a reasonable amount of technical information included that may not be necessary to understanding the thesis. However, I’ve included it in that it’s critical for proving the sometimes difficult case in technology stocks of why there are high barriers to entry and sustainable competitive advantages.

Ion Implantation is a 1.4 billion sub sector of the semiconductor equipment space, historically representing 3-6% of the overall market. Ion implantation is a low-temperature vacuum technology that uses a linear accelerator to create a beam of charged atoms, or ions. The ion beam is then shaped and directed toward the target surface, embedding metallic ions (such as boron, arsenic and phosphorous) into the material such as silicon to change its chemical, physical, and/or electrical properties. It is one of the key stages of the front end of the fabrication process where important components of the chip such as the source and drain regions and gate channel, are formed. The doping process is what is largely responsible for the unique electrical properties of semiconductors. A good analogy is it is like a spray paint can which you spray onto a surface with the aid of a stencil, only the particles are specially selected to eliminate contamination. The machines weigh about 25 tons and take up the space of a garage, selling for $3-5 million ASP.

What is attractive here unlike many capital intensive equipment businesses is the fact that the continual need for more advanced replacements prevents commoditization. Semiconductor production is a rapidly evolving business, where the wafer size changes approximately once every decade now (from 200mm to the current 300mm) and the chip size changes every 12-18 months (from 90nm to 65nm to 45nm today), so constant reinvestment in equipment is necessary. Ion implantation demand in particular has benefited because the added sophistication at smaller chip sizes has required a rapid increase in the implant steps (up to 35 today, compared to 6-8 in the 70s). It has become much more difficult to achieve high yields and throughput. As a result, fabs are turning to implanters not just for capacity but also for productivity gains as it becomes a more challenging step in the process, increasing the value proposition and pricing power.

Ion implantation is separated into 4 segments that each have distinctive competitive dynamics. Ion implantation devices can differ in their capable range of energy and dose levels. The combination of these different machines with their ranges of capable energy and dose levels creates the ability to perform all the processes required.

The high current implanter is designed to be low energy and high dose. It is used to fabricate the source and drain regions, which are some of the smallest regions of the device being fabricated (equal to the size of the gate length). This is why it must operate at low energies to achieve shallow implantation. It also needs high dopant levels because the source and drain act as electrical contacts. As the depths of the smallest parts of the semiconductor shrink, drains and particularly the source drain extension, a small nook next to the gate, become increasingly shallow. This drives the need for higher dosing and lower energy.

High current is the largest segment of the ion implantation market, and historically half of the total market. It has traditionally been highly competitive due to added competition from AMAT. The split of batch and single wafer technology also kept any player from gaining a dominant position.

The medium current implanter operates on dose and energy levels in between the high energy and high current/plasma segments. It is a more flexible machine that is additionally attractive because it has more control such that it can be used to do some things a high current machine can also do in cases where precision is less important. This implanter is primarily used to implant the gate channel because it needs neither high doping for very high electrical uses or the deep penetration of high energy.

Medium current has always been a single wafer technology due to the implant angle requirements. Varian has historically held the majority market share, the only significant other player being a Japanese company, Nissin Electric, a medium current specialist. Nissin Electric has typically held the large majority of its roughly 25-30% market share from Japan, where Varian had a very minimal foothold. Now as Varian is starting to get significant traction in Japan, it should start to pick up a much larger share of the high current market.

The high energy implanter tool is used to dope the wells and isolation regions of the transistors as they lie deep in the wafer substrate; higher amounts of energy are needed to penetrate to that depth without the need for high dose. High energy has experienced little to no growth in recent years, as the need for it declines as technology scales.

High energy is the only segment that is still operating primarily on batch wafer technology, which Axcelis specializes, hence their market dominance. A shift to single wafer over the next two years will open up the field for Varian, disrupting Axcelis’ incumbency edge. Varian has the added advantage of synergies with its common line platform and a product series that has been vetted in the market longer, since Axcelis just launched their high energy single wafer tool in November 2007. High energy is becoming a battleground for market share despite its small market size (18% of implanter market), lower margins and slow growth rate because it is incremental revenue for Varian, which will give them a monopolistic share of the overall market and take market share from Axcelis’s core market.

The first commercially viable plasma tool, Varian’s PLAD, is currently used by memory manufacturers specifically for a new application called dual poly gate, previously done through disposition. Thus, it is at the moment building a new market and not cannibalizing demand for high current tools. Unlike ion implantation, where ions are shot onto the wafer, the gel-like plasma is placed on top of the wafer and sucked into wafer from below. This means that the tool can achieve an “ultra high dose” or current but conversely the energy is low, meaning the implantation is very shallow. These qualities expand the scope of what implantation can accomplish, and the possibilities for this new technology are potentially huge.

Varian has been developing it for over a decade, testing manufacturers in 2005 and last year they began recognizing revenue, a total of 47 million in 2007. PLAD operates with even lower energy and an ultra high dose compared to high current. Filling a growing technical gap it is a very attractive new application that is expected to conservatively reach sales of 120 million in 2009 despite the current downturn.

There are many other applications Varian is working on qualifying PLAD for that should greatly expand its role in the fabrication process, helping expand the TAM. Additionally, ion implantation but more specifically plasma has the potential to extend the practical uses of implant technology beyond semiconductors for commercial purposes although this is not yet the focus. Areas of interest in implementing the technology include anti-corrosive applications and solar technology.

The main driver of change in ion implantation over the past 5 years was the shift of high current from predominately batch ion implantation to single wafer. Single wafer technology involves placing a single wafer on the implanter and doping one wafer at a time. This is a slower process but it has advantages for the tilt of implanting (zero degrees can be achieved) and contamination issues, which have become increasingly important at smaller nodes. Batch involves putting a lot of wafers on a pinwheel like device that spins as the wafers are implanted, which is much faster. A problem with batch technology is that the contamination issues became a serious issue by the 90nm node. Pinpoint accuracy became necessary and since batch was less accurate, the minimum implant angle was 3 to 7 degrees creating the issue of shadowing; at the same time, single wafer have been able to close the productivity gap as the move from 200mm to 300mm wafers increased the efficiency of the single wafer process. The quality advantage of single wafer quickly outweighed the quantity advantage since batch no longer produced competitive throughputs due to these contamination issues that affected overall yields.

Varian was correctly positioned to benefit from this trend since it had been using a complete line of single wafer products that had been its research and production focus for years. Conversely, competitors Axcelis and AMAT had yet to even test their single wafer products as the shift was starting in 2004-2005. Furthermore, Varian had certain patents on advantageous single wafer technology (Axcelis had still been betting on batch through 2004) such as dual-magnet ribbon technology, which helps filter out particles in the ion accelerator creating the cleanest ion beam. They also are the only implanter maker able to achieve a zero degree angle which prevents the shadowing problem. Its VIISta integrated platform offered an easy change, mitigating switching costs since most customers already used its medium current devices. Competitors weren’t just late to compete but had inferior technology partially due to not focusing on single wafer technology.

There are 4 parts to an implanter, the software, tool, controls and end station. Through Varian’s integrated VIISta platform, the controls and end station are uniform for every tool Varian produces. This is something no competitor does and benefits Varian enormously in two ways. First it confers large productions cost savings through economies of scale and better product reliability, which lowers warrantee costs Secondly, it streamlines R&D, again providing R&D economies of scale for the controls and end station and minimizing the research necessary to develop upgrades to their product lines. As a result, Varian has been introducing new products every 12-18 months compared to every 3-4 years for competitors. These rapid upgrades also drive profitability because they can (and always are) sold to the existing VIISta line tool base for a couple hundred thousand dollars; since they are mainly done in the form of software and have very high margins.

The ion implantation market is a good example of an industry with dynamic change but (until recently) a stable competitive environment with massive barriers to entry, and a true oligopoly of 3. Applied Materials, Axcelis and Varian over the years have formed 80-90% of the marketplace with the vast majority of the remainder coming from the protectionist Japan’s SEN and Nissin Electric.

There have been no meaningful entrants into the business in decades, with all of the big 3 tracing their roots in the business back to the 70’s, when mass production and commercialization of semiconductors took off. The reason for this is the high degree of intellectual property these companies possess, and just as if not more importantly, the technical experience/human capital in developing and producing such intricate and unique machines creates a huge barrier. It is not a coincidence that every ion implanter today except Nissin Electric and over 90% of the market derives its technology from a 20 mile area north of Boston (the industry has been called the analog Swiss watchmakers of the semiconductor industry). Even though it is an attractive, good margin business, the size of the market opportunity has never justified the difficulty and startup costs of trying to take on two specialists and the largest company in the industry. These barriers have grown over time as the business has become more complex with chip size shrinking and the difficulty of developing adequate solutions much increased.

Varian’s customers are risk adverse given the huge capital outlays of a new fabrication plant that can amount to $2-3 billion, where ion implantation spending is only $100 million. When a fab has chosen a PTOR (process tool of record), they are unlikely to change unless the differences are significant. . Even the decision to have a competing secondary tool may be overruled because having the best product is just too vital, which is currently what is taking place between Varian and Axcelis in the high current segment. Varian has become the PTOR at all of the top 20 companies outside of Japan for high current and nearly all of the medium current. As discussed previously, they are making inroads into Japan and high energy which should start to show in FY09.

Varian has been ranked #1 for 9 of the last 10 years in the customer satisfaction survey for large suppliers of semiconductor equipment by VLSI Research, a leading semiconductor think tank. The 2007 award included this statement:

“This award was based on customer survey results from more than 570 individuals at 48 different companies worldwide. Varian Semiconductor received the best ratings among its chip making equipment peers in the following categories that are critical to customers: cost of ownership, uptime, software, build quality, usable throughput, quality of results, product performance, process support, field engineering support, spares support, support after sales, and commitment”

Varian’s management has a history of good execution, sandbagging, and strong experience. CEO Gary Dickenson, age 50, joined Varian in late 2004, and was previously president and COO of KLAC for the two years prior. CFO Bob Halliday joined Varian in 2001 in his current role and has previous CFO experience at smaller technology companies.

Applied Materials operated in the high current segment, possessing the advantage of cross-selling along with strong relationships with all the chip manufacturers. This allowed them to become and retain PTOR relationships for decades and in most years have the largest market share. However, AMAT’s high current tools were batch and they were late to develop and bring a single wafer tool to market due to the long lead time in developing and getting acceptance for a new product line. As a result, ion implanters for AMAT performed poorly in 2005 and 2006 as their new single wafer tool, the Quantum X,struggled to gain traction. Varian was thus able to make inroads into some of AMAT’s 2 decade old customer base.

After considering all alternatives, including shopping for a buyer, AMAT surprised industry experts by exiting the high current market in February 2007. AMAT said they didn’t see future R&D returns justifying remaining in a business becoming commoditized. However, this wasn't the case as Varian, who in reality had forced them out, was seeing increasing margins due to high value-added and thus less commoditization. Varian recently finished absorbing all of AMAT’s roughly 25% remaining high current share.

AMAT has been experimenting with plasma technology for implantation for two decades but nothing has come of it. Their P3i tool has gained very limited traction so far, shipping 4 tools and booking 1 with Samsung (who has been most aggressive in building out plasma tools with roughly 15 PLAD tools sold). Indications point to a limited role in the market or a possible exit altogether.

Although in a woeful competitive position, Axcelis remains Varian's largest threat. Located next door in Beverley, MA, Axcelis was founded 7 years after Varian in 1978 by former employees. Until recently it closely rivaled Varian, producing in all three segments and over the last 3 decades vying in a close race for top market share (though Varian always held the largest installed base).

However, Axcelis has been poorly run in recent years by CEO Mary Puma. By the time they released their single wafer high current tool, the Optima HD, it was already delayed a year and most companies had already switched to Varian’s high current product line. As a result, the Optima HD has experienced very weak traction, with about 15 tools shipped (5 purchased, 10 under evaluation) in the 18 months since the launch (less than Varian high current sales in any quarter) and of those, the 5 purchased will be finally recognized as revenue next quarter. Despite these new tools, Axcelis has still guided down Q/Q and had a 30% Y/Y decline in Q1. At this point it would be upside to Axcelis’ share price if the Optima HD proves just reasonably competitive and can make some headway as a viable secondary tool.

Of their 400 million in revenue in FY07, about 40% was tied to legacy service revenue from their installed base and another 18% came from non Ion Implantation products (dry strip). Only $155 million came from new implanter sales, and $120 million of these were in high energy. If the stake in SEN was rolled into Axcelis’ top line total implanter sales would have been about $250 million. Axcelis is facing cash burn of roughly $20 million/year and margins about 15% below Varian’s, caused by poor product mix, cost structure and economies of scale ect. Most importantly, Axcelis is facing an inability to match Varian in R&D spend ($145 million vs $200 million for Varian in the last 2 years), let alone the efficiency of that spending.

The situation got more interesting in February when Sumitomo Heavy Industries (SHI) made an unsolicited offer of $5.20/share for ACLS representing a 20% premium to the current price, although a 10% discount to the 52wk trading average. ACLS had gotten as low as TBV of $400 million recently. Axcelis has been the target of merger speculation for over a year given this base valuation and its strategic position as the only alternative to Varian. SHI though is a unique acquirer because of the strategic connection through its 50-50 JV with Axcelis. This was formed with the purpose of being a way for Axcelis to get into the protectionist and sticky Japan market. SEN sells Axcelis’ equipment solely in Japan, paying a royalty as well as the income split from the JV. In recent years, SEN and Axcelis have been in royalty disputes.

Axcelis, as expected, rejected the lowball bid this week These companies are too much in bed and Axcelis is too disorganized for the merger not to take place at a modestly higher price. With Axcelis trading up 6% to 5.75 on a very weak tape on 2/29, the market is suggesting this to be the case. However, given the all circumstances, the shareholder lawsuit just filed, it will be messy and disruptive short term and should only further work to Varian’s favor. It will inevitably bring new leadership and capital resources to Axcelis but the technological gap is already so substantial Varian’s entrenchment as PTOR and the length of product cycles means that if Axcelis can turn around the ship at best this would make them competitive with Varian no earlier than 2010-2011. Quite possibly they could eventually move to exit the high current market like AMAT and just remain in high energy and with their large service revenues.

There are a few fringe competitors. Nissan Electric operates in the medium current, the only other player of significance in that sector. The second is Sumitomo Eaton Nova (SEN), a 50-50 joint venture formed by Axcelis and Sumitomo Heavy Industry in 1993. Axcelis licenses technology to them and therefore their segment focus correspondingly mimics Axcelis. Collectively, including the half of SEN owned by Axcelis, they have about 20% of market share today, down from 27% in 2006 and 13% if you exclude Axcelis ownership. There are a couple other competitors such as Semequip which licenses a cluster beam product to Axcelis and Ulvac, in the Boston north shore with less than 1% market collectively.

I predict SEN and Nissan's market share will shrink since the segments they are not in are the higher growth areas, namely the non-Japanese semis equipment market (Japanese capex now 19% from 28%) and PLAD. They should face more competition from Varian as it makes inroads in the protectionist Japanese market. The increasing competition from Varian finally on a more level playing field should pose serious issues for SEN and Nissin. Over the next few years their market share (not included the Axcelis share) should decline to high single digits.

Financials and Valuation

2008 is going to be a very tough year for semiconductors, especially in DRAM and their suppliers, by all estimates 10-20% down Ion implantation is more exposed to the weakness in memory but balancing this is stronger sub-sector trends as well as the adoption of the new high growth segment, PLAD. These should help balance out over exposure to memory, yet to be conservative I’m forecasting an 18% decline for the industry excluding PLAD in FY08. Further, I’m forecasting 3% growth excluding PLAD in FY09, which likely leaves upside should the recovery start before the 2^nd half of 2009. I feel this is possible as semiconductor cycles have been moderating. From there, the industry growth rates assume a moderate two year recovery followed by industry maturation outside of plasma. Overall assumptions arrive at a 5 year industry CAGR of about 5% driven by plasma expanding the market (12% CAGR for Varian). I believe this may prove cautious given that it assumes negligible growth outside of plasma, which has not been the trend before the downturn.

Market share assumptions for FY08 assumes a few wins in Japan, the absorption of remaining AMAT business, and some customer adoption of their single wafer high energy tools. FY09-10 assumes further penetration into Japan and more broad based wins in high energy as the shift to single wafer intensifies.The reason it seems very likely Varian can achieve sustainable 80-85% market share is this is what they already have in the non high energy and Japan implantation market. As previously discussed, these are areas where there is significant evidence they will have capitalize on the success they have been having in these markets. The smallest and shrinking markets of the industry, high energy and Japan, are expected to continue their decline in relation to the overall market.

I forecast a moderate progression in margins from about 48% in Q1 2008 up to sustainable levels in the low to mid 50’s in FY11. Margins rose 600bps for 2007 YoY. This was achieved while Varian made huge market share gains, which initially come at the cost of margins because of the additional costs to setup and support machines with new customers and the lack of Service Spares and Upgrades, which have higher margins and only kick in after about 15 months. Thus, despite what should be a temporary drag on margins, a systemic change to gross margins more than compensated, proving how dramatic margins are improving. Specifically 200bps of the improvement are attributed to the VIISta scale and other production efficiencies; management expects to achieve another 200bps from cost savings going forward. Other incremental continuing trends are mix improvements from SSU’s and PLAD, which will offset the negative mix of growing lower margin high energy sales.

Operating costs should benefit from economies of scale, particularly R&D. SG&A should decline slightly over time from 22% towards 20% as Varian scales. A big impact on income will come from Varian’s declining tax rate. Varian took a large onetime tax hit (49%) in FY07 to move its IP to Switzerland. As a result, management is projecting the tax-rate to decline over time towards the low twenties-high teens. I have modeled a logarithmically declining rate from 32% in FY08 to 24% in FY13 to be conservative.

Varian’s FCF is growing nearly in line with earnings, growing at slightly under 30%/year as capex and working capital needs have been consistently minimal. Varian’s production comes all from its facility at the headquarters in Gloucester MA. In the last two years, Varian has used its nearly half billion in cash to repurchase shares. I expect this to continue as there is no viable acquisition and they have no intention of paying a dividend.

Varian will continue to do massive buybacks and likely cut shares outstanding by 30+% over six years. The buybacks have been modeled in by assuming an average buyback price every year and using a % of cash left available added to that year’s FCF. In the last year of the projection, two scenarios have been created. In the first they continue with a straight buyback and in the other they adjust the capital structure adding about a billion of debt with would still leave them with interest coverage of over 12 to 1. In the debt case, the terminal value discount rate is 1% lower.

Varian has a projected 5 year revenue CAGR of 12% and, as a result of the expansion in profit margins from 10% in FY06 to 20+% and the positive leverage of the share buybacks, a projected 5 year EPS CAGR of slightly below 30%. A DCF model, using assumptions of a 15% discount rate (14% for the terminal value of the recapitalization scenario) and a 2% growth rate, arrived at 47 and 52 for the straight valuation and the recapitalization respectively. Discounting back 2013 earnings of 8.42 at 15% and applying a mid teens multiple leads to a valuation in the same range of about 60. Therefore, I believe Varian from current levels could be a double. Given the acceleration of earnings and secular growth story, such a valuation should be realized within 2-3 years with rather limited downside from levels in the low 30’s.