|Shares Out. (in M):||0||P/E|
|Market Cap (in M):||788||P/FCF|
|Net Debt (in M):||0||EBIT||0||0|
Cymer is the technology and market share leader at “the bottleneck” in path forward for semiconductor performance, and is showing signs of being about to distance itself from its one and only competitor.
After aggressively completing a $300MM share BB in 2007, CYMI ended the year with $357MM in cash & eq and only $140MM in debt ( 3.5% convert @ $50, due 2/15/09). It generates over $25MM FCF/Q.
Unlike in past market downturns, over half of its revenues now come from spares, upgrades and other non-system sales to support an installed base of over 3250 light source systems.
Cymer is the clear leader in the design and production of light sources for photo lithography used in semiconductor wafer production. These light sources are used in scanners and steppers produced by ASML, Nikon and Canon. (Photolithography is similar to film development. The silicon wafer is polished and then light sensitive material is applied. It is then masked and exposed to UV light, which alters the material that was not masked. The mask is removed, more chemicals applied to remove materials and leave the desired pattern. This is repeated hundreds of times to create between 28 and 43 layers for most microprocessors.) As circuitry node sizes have decreased, more precise light pulses (wavelength, wattage, duration, among other measures) are required. Cymer emerged as the leading producer of light sources based on its understanding of excimer lasers, leading the way from a krypton-fluoride (KrFl) based process to a dual-chamber, argon-fluoride (ArFl) system. With the demise a few years ago of Lambda-Physik, only one competitor remains. GigaPhoton is a JV between Komatsu and another Japanese company. As chip makers look to advance from the 65nm to 45nm node size and beyond, the industry is counting on CYMI to produce the light sources that can deliver to evermore exacting tolerances and at lower total operating costs. CYMI has made initial production shipments of ring-based “immersion” technology, which will extend ArFl technology to the 32nm node size. The next step will be accomplished with a technique called double-patterning. Beyond that, the likely final step will be Extreme Ultra-violet (EUV). CYMI has been working on EUV since 1997 and in late 2007 received initial orders from ASML for EUV light sources. This is a steep technological curve, and it has been rumored that Komatsu is not sure it wants to fund spending to keep following CYMI.
Besides delivering leading edge technology for the critical layers on next generation wafers, CYMI also disclosed in its latest conference call that they have re-tooled some of their older KrFl product, where their competitor has been able to gain some market share with aggressive pricing (customers will always support a reasonably competent second source). Many of the layers in microprocessors can be done with these non leading edge technologies. It is likely that while KrFl is no longer leading edge, it can be enhanced to take share from i-line systems (the technology that preceded KrFl) for many of these non-critical layers.
In addition to what appears to be a widening lead on the new product front, Cymer is in the midst of re-engineering its aftermarket offering to a model that is more fixed price. Called OnPulse, it will provide 24/7 support based on the number of pulses the customer gets from the equipment. As photolithography is very messy as things go in the fab world, light sources generate relatively high parts & service revenues over their typically very long lives. It is management’s expectation that OnPulse will move the margins on this part of the business, which has been somewhat below the corporate average, up to that present average.
Revenue Growth? CYMI’s revenue growth correlates with the growth in demand for wafer fab equipment (WFE), but when you look at units and ASP it gets complicated. Over time, the productivity improvements that CYMI has been able to introduce has translated into a downward trend in the number of light sources but an even more powerful uptrend in ASP. This has been described in terms of “If a KrF light source was priced at 1X, think of the ArFl we introduced a few years ago as 2X, the immersion systems as 3X, and double-patterning systems as 4X.” ASP grew at a 9% CAGR 2002-2006. Lately, the initial shipments of immersion (as opposed to dry) ArF systems is pushing ASP up (25% in 2007) while restrained capex is weighing on non-leading edge (KrF and i-line) lithography spend. These trends are expected to persist well into 2008. For Units, I started out assuming that they maintain a 65% share of forecasted demand. (The industry unit forecast is flattish, 370 in 2011 v 355 in 2007.) Cymer’s market share is more like 90% at the high end (immersion ArF at $1.6MM/unit) and 50% in KrFl where the competitor can compete on price. They have addressed share loss at the low end with a product refresh first mentioned on the Q3 07 conference call. The next “bump” in ASP will be double patterning light sources, which is expected to ship in 2009.
I believe my revenue assumptions, which grow revenues from $526MM in 2008 to $682MM in 2012, are conservative because Cymer is so dominant at the high end of the price and they have a credible game plan for regaining unit share at the low end. (Offer a product that can so obviously produce more through-put with less space, power and/or downtime that you can charge a premium for it.)
Revenue growth also comes from growth in consumables & spares for the installed base. This revenue has grown at a 20%+ CAGR since 2002. I am assuming high single-digit growth going forward. CYMI is changing its aftermarket model with OnPulse. It sounds like an improved value proposition for the customers and if CYMI correctly estimates the improved durability they are engineering into their systems it will be more profitable for them to charge what amounts to a fixed rate for uptime.
Earnings Outlook 2008 is looking to be a down year for CYMI. The consensus estimate is presently $1.91, down from $2.50 in 2007. This is due in part the slowing in WFE spending that first became apparent in the industry in Q2 07 but did not register for CYMI until the very end of the year. Due to certain customers’ need for next generation “immersion” systems, the company is confident that H2 will be stronger than H1. This pause in demand will cause reduced fixed cost absorption and so a lower GM (48ish% v. 50%ish) in H1. Additionally, the company has stepped up spending to support the roll-out of OnPulse, and R&D spending, which totaled $81MM last year, or $2.67/share, is expected to trend from just under 15% of sales to almost 18%. Taking all this together, I am expecting the operating margin to decline from 22.8% to 15.5%, and EPS to decline to about $1.80.
Once the slowdown that started almost a year ago has run its course and demand for next-gen photolithography kicks in, I am looking for moderate volume improvement to lead to better leveraging of costs. R&D will continue to grow, but at a much slower rate. I think it is reasonable to expect the operating margin to improve to 21% in 2009 and then trend higher, getting just above 23% in 2012. Below the operating line, we have to consider either an increase in net interest expense as the notes go away in early 2009 and the cash builds, or another reduction in the share count. The $300MM repurchased in 2007 followed $50MM and $100MM in the prior two years, as well as $57MM redemption of the notes.
Appreciation Potential? This flexibility in the use of surplus cash presents a challenge in arriving at a likely future valuation based on EPS. At its last peak in 2006, CYMI sold at $56.70, or 23.7 times earnings. At present, it is selling for a little over ten times last year’s earnings despite having net cash of over $7/share. To get around the conundrum of not knowing what the EPS denominator is going to be, I would look at prospective EV/EBITDA assuming the note gets paid and the cash just piles up. EV/EBITDA was 12.29 at the 2006 peak. If we apply that multiple to 2008, AP is $54. If we apply it to my projection for 2011, it is $95. If we haircut this by 20% to 9.83 times EBITDA to allow for how things might change in the next 3-4 years, we still get AP to about $81.
In allowing for such contingencies, though, it must also be pointed out that I have not said much about what could go really right for CYMI in terms of the competitor falling further behind or even giving up, or the possibility that they manage the eventual “end of the road for excimer lasers for the critical layers” astutely. (I should also mention that they have a potentially interesting JV with Zeiss called TCZ, that has developed tools for the LCD flat panel industry, and is expected to ship its first production tools in H1 08. I have not figured this into earnings outlook.) I recently had the opportunity to attend Intel’s Investor Day and any fears I might have had that “node shrink” is about to reach the end of the line have been put to rest. Even after we get to where photolithography has reached its limits (I am thinking 6-8 years from now) there will be demand for a couple of decades at least based on serving the installed base and continued unit demand for technology where non-critical layers are done with CYMI light sources. If they recognize “what time it is”, they could actually grow earnings by scaling back R&D and otherwise adjusting their overhead structure. There is the risk that they will go “a bridge too far”, stepping up R&D for a couple of years only to learn that their technology is no longer capable of going to the next level. This is not likely a concern today, but it is something to keep an eye on longer term.
The Bear Case There is a bear case going around on CYMI that merits some attention. It revolves around market share loss to GigaPhoton, their Japanese JV competitor. As is often the case, there is a kernel of truth here, but it has been blown out of proportion. When CYMI lead the way from KrFl to ArFl a few years ago, the competitor had a hard time following, and so for a season CYMI had the then-leading edge all to itself. Three years later, the reverse engineering has succeeded, CYMI is focused on moving the leading edge to where the competitor might not even want to go, and not surprisingly, the competitor has qualified some product and is taking market share in unit terms. (You get a different market share picture in dollars, as CYMI maintains a 90% share in the light sources that go for multiples of what the technology that hit the market five years ago goes for.) There is almost no information available from Giga-photon, as they are a JV, but there seem to be sell-side analysts who express grave concerns because their “sources” tell them that Gigaphoton is taking share. CYMI is clearly acting in a way that will address the competitive threat at both the leading and trailing edge of the product mix, but they are adamant that they are not losing share in any sense that really matters. I am more inclined than usual to believe this, because if anyone knows about market share, it’s them. Their equipment is in every fab in the world and their people visit on a regular basis and see what is going on. Giga-photon cannot say this, no customer can say this, and certainly no sell-side analyst has “sources” who can say this (with a straight face). We should expect that as in any business, customers want to avoid depending on a sole source and will do what they can to help a second source along. There is some possibility that CYMI’s profitability benefited for a period from an effective monopoly on something its customers needed badly, which would mean that earnings on that product will revert from that unnatural level as the alternative becomes available, but this will have little bearing on how leading edge equipment and replacement parts for the 3250+ CYMI light sources that are in place get priced over the time period we are concerned with.
I recommend CYMI at this time because of its leadership position enabling the likes of Intel and Samsung to get where they want to go, its impressive financial flexibility, its minimal downside risk and an odds-on likelihood of reaching or exceeding its 2006 peak price of $57 sometime over the next 2-3 years.
|Entry||04/14/2008 05:40 PM|
|Interesting idea. One thing that scares me about this one besides the rumors of market share loss is that from what I can see 2008 revenue is currently estimated to only be down 3%. Do you think this is realistic or do you think estimates will come down further as the year goes on? From a high level view I know that semiconductor capital equipment spending is supposed to be down more than that, and as the economy looks increasingly weak and memory prices are in a freefall it seems like there is a risk that spending will fall even further than expected. I know these stocks usually bottom before the estimates do, but lately it seems like no matter how beaten up a stock is it keeps going down if revenue and earnings estimates are coming down.|
Funny enough you posted this right before CSFB downgraded to underperform based on the bear case you laid out. I take it from your description there was nothing in the report that you didn't already address. I don't find it as farfetched as you do that some of these analysts could have good sources, there are only a handful of lithography vendors out there and many analysts cultivate contacts with them since lithography is the highest lead time piece of equipment.
|Entry||04/23/2008 09:40 PM|
|Quarter was slightly better than I expected, but real news was that the aftermarket initiative is finally showing some momentum and mgmt confirmed that while the slow rate of spending by customers will persist thru q2 there should be a significant pickup beyond that. Customers reduced WIP by 24 units during Q to 81, a level that cannot go much lower. Spares & upgrades sales were surprisingly strong, reflecting a new upgrade to better manage gas (GLX) and the fact that customers have pushed the equipment pretty hard trying to get more out of what they have instead of adding equipment during a downturn that started almost a year ago. OnPulse agreements have been signed by a number of customers. This seems to have been one of those initiatives where "nobody wants to go first" but now they are talking in terms of having 800+ of their installed base of 3300 covered by this fixed-rate, pay for update program by the end of Q2. Margins should benefit by YE. This new commitment, which is facilitated by their ability to remotely monitor performance in customer sites, has required additional inventory to be on hand in the field. THis, and the rollout of immersion systems, is the reason mgmt gave for the increase in inventories. |
Mgmt also stated that their share rose during the Q to 72% using their rolling 4q methodology. As I stated in my rec, they are better positioned than anyone to know about share of such a large diverse mkt. Having once been a SS analyst (diffent industry) I know value of cultivating "sources", but I also know the limitations. It would be silly to suppose that customers will not do what they can to foster second sources in any business, and when your share is so high it will eventually go down if you are standing still. WHat I see here, though, is a company that is not standing still, and the road to productive EUV, which will see pilot deliveries next year, will be much tougher to follow.
|Subject||Update following Analyst Day|
|Entry||09/19/2008 11:38 AM|
|After attending the recent Analyst Day, it seems like a good idea to make a couple of updates to my initial recommendation. The most important one is that my initial assessment drastically underestimated the positive impact that CYMI’s growing installed base is starting to have on the level and stability of sales and earnings. The other, less significant one would be that the downturn semiconductor equipment capital spending cycle, which peaked in Q2 07, continues to “move to the right”. Only a few weeks ago, industry leaders like AMAT were speaking confidently as if Q3 would be the trough. Fear and loathing emanating out to the financial markets has translated into uncertainty that has started to be felt in industrial production rates, including DUV photolithography scanners. Not dramatically, but enough to slow down a decision about when to pull the lever on the next de-bottlenecking or expansion. It has not helped that the largest (43% of demand) segment, Memory, had built capacity to support growth in solid state drives, but for various reasons and despite all the hype, 80GB SSD priced at $500+ just aren’t making a dent against $80, 500GB hard disk drives. Memory makers will, one way or another, grow into their excess capacity, but it won’t happen as soon as anyone was thinking several months ago, a year into the current capex downturn. |
The most significant insight I gleaned from my day with the company was the evolution that has occurred in their aftermarket opportunity. Installed base parts & service (IBPS) was only about $80MM in 2000 and $100MM in 2003, but is on track to hit $330MM this year. The primary driver of IBPS is the number of pulses generated by the 3300+ systems in the installed base. It was obvious that there is a great aftermarket opportunity here, we are talking about gas-generated lasers bouncing off of mirrors billions of times, but the lumpiness of the annual spending trends masked the overall growth rate. Pulse count has grown by 3.5X over the past five years. It is partly the number of systems, but it even more so the greater productivity of the argon fluoride (ArF) systems that started entering service in 2003 (2-3X the number of pulses as the krypton fluoride systems that were the previous state of the art). It also matters that an ArF system can be expected to generate aftermarket demand on the order of 4-6 times its $1.5MM+ selling price over its working life, as opposed to 2-3 times that of a $500K KrF system. CYMI recently estimated that pulse count was growing (during H1) at about 2-3% per month, (2.1T pulses/month, system-wide, 50% higher than two years ago). At the Analyst Day they said that there will likely be some slowing in H2 08, but the longer term growth rate in the highly energetic activity that drives replacement part demand is sufficient to support double digit growth in IBPS revenues, year in, year out.
I also learned a bit more about On Pulse, the new business model for IBPS. The company was optimistic about it back in April but had nothing to say about customer adoption rates. There seemed to be a sense of “everyone waiting for someone else to go first”. It turns out that the customers are going for it in a big way. More 950 systems are now under On Pulse agreements. One thing that I did not fully appreciate six months ago is how secretive the customers in this industry are about their operating data (relative to other industries I have known). At the meeting, Cymer disclosed how they have started to do remote monitoring of customer systems. I asked why something I was hearing about in other industries at least five years ago was now only starting to happen. Apparently, CYMI has achieved a level of trust that they will safeguard the confidential nature of info that pertains to “who is doing what at what rate, etc”. This remote monitoring has enabled them to develop systems for predictive diagnostics, which among thing is going to make On Pulse a great differentiator (v. the one & competitor), enabling them to deliver improved up-time in way that they and the customers share in the savings.
Under On Pulse, the customer pays a fee on an annual basis that fixes their cost per pulse. The formula gets adjusted each year in a way that reduces the cost per pulse (the number of pulses increases, but the cost of coverage increases less). This could be viewed as risky, except that CYMI has such a good understanding of how much it should cost to keep the systems operating. They also have developed enhancements that improve productivity. For example, over the past two quarters they have introduced GLX, a gas life extension module. A second generation, GLX2, extends the time between a downtime between system gas replenishments by 20X. From a customer point of view, that’s an additional seven hours of uptime, or an estimated 470 wafers per month. With the recently added ability to sit in their TacOps center in California and monitor how the systems are performing, they can optimize the field service effort (presently 165 field service engineers in 44 depots supporting 110 customers). In effect, this is outsourcing, as customers are able to offload a meaningful slice of the complexity that is operating a fab. On Pulse will not only improve the margins of the IBPS part of the business (presently almost 70% of revenues, it should be 60%+ of the total most of the time going forward). It will also engender an unprecedented level of customer intimacy, one more thing that the competitor, who lets the scanner maker handle the aftermarket, cannot possible match them on.
Not surprisingly, the earnings recovery will be slower than I expected in April, but CYMI will remain solidly profitable. Look for EPS on the order of $1.30 this year and somewhat higher in 2009 unless we are headed into a global recession that derails demand for all things using microprocessor control. Assuming no meaningful share repurchases, my model projects EPS on the order of $3 in 2010 and $4 in 2012. Importantly, because so much of those earnings are coming from an installed base that continues to grow, the market should eventually figure out that CYMI’s earnings are not as cyclical as they used to be and reflect this realization in a higher multiple. I also believe that as Cymer would have over $5 per share in cash if they paid off their only debt, a 3.5%, $140MM convertible (@ $50) that is due in February, they will at some point get aggressive about share repurchase again. (They bought $450MM worth 2005-07, but as of August on $22MM on their latest, $100MM authorization.)
Having suffered through the bottoms on more than a few cyclical stocks over the years, I am not surprised that the Street coverage of CYMI is as “battened down” and paralyzed by fear as it is. That experience has also taught me, though, that as hard as it is to see the next peak from the bottom, it is reasonable to wonder whether the return on CYMI is going 2X, 3X or even 4X off its recent low over the next 3-4 years.