After a period of "brute force" capacity expansion, the photonics industry is now entering a demand-driven growth phase that will see the development of a mature supply chain, according to Eric Chen and Donald Lu of J P Morgan H&Q Equity Research.

The photonics sector is destined to become a major force in the world economy. Although the market demand for photonics components and systems is potentially massive, it is our view that the current manufacturing approaches employed in the photonics business are too inefficient for the industry to address all of the available markets. Contrasting the photonics supply chain to the infrastructure of "Silicon Valley" yields some useful insights. In the silicon semiconductor industry, device manufacturers rely heavily on the suppliers of equipment, design tools, and even pure intellectual properties. Together, these companies form the holistic semiconductor supply chain. For investors there are Intel, which makes semiconductor devices, and Applied Materials, which provides semiconductor capital equipment. Both companies can be suitable investment vehicles to cash in on the opportunities in semiconductors. On the photonics side, the picture is far from clear. The technology supply chain remains underdeveloped. If JDS Uniphase is poised to become the Intel of photonics, where is the Applied Materials of photonics? Analyzing this question is important, and not just for companies such as ours that are always looking for great investment vehicles. It is also important for the overall health of the photonics industry. Recipe for disaster We believe that the current processes employed in photonic device manufacturing are prohibitively inefficient to support a profitable growth business over the long run. Take the industry leader, JDS Uniphase, as the representative company, and assume that JDSU s manufacturing economics are typical for a photonics manufacturer. Each employee at JDSU generates roughly $110 000 of revenue per year. Considering the mix of low-cost labor employed by these companies, this can be a reasonably profitable business model. It suggests that there are approximately 125 000 people involved in making photonics communication products in the world today. Now consider the future. The market research firm RHK estimates the market for fiber-optic components will grow 35% per year until 2003. We will extend this forecast by assuming a 25% growth rate after 2003, and we will further assume that the component average selling price (ASP) will decrease 25% per year. Under this scenario the current model of employee productivity says that nearly 34 million people would be needed in the photonics business by the year 2011 (more than the entire population of Canada!), and each employee would generate only $4000 of revenue a year. Clearly this is an unsustainable business model. The only way to fix the model is to change one of the assumptions: either unit and ASP growth rates for devices, or the manufacturing economics. We believe the latter is more likely to change than the former. Again, the silicon industry provides an example. In 1983 Intel generated $52 000 in revenue per employee. Sixteen years later in 1999, Intel had increased its revenue per worker to $420 000 (see ). This was done despite the fact that ASPs decreased 25% per year over the same period. The main driver for this remarkable improvement in productivity has been the relentless effort of the semiconductor industry to increase wafer size, to improve yield, and to implement highly automated manufacturing processes. The photonics industry pales in comparison, a direct reflection of the highly labor-intensive process of photonics manufacturing. For an example drawn from the compound semiconductor industry, look at the difference between Vitesse and SDL. While Vitesse has almost doubled its revenue productivity in the last four years, achieving "Intel-like" numbers, SDL has seen basically flat revenue per employee. During this period, both companies greatly improved the yield and productivity of their III-V semiconductor operation. In addition, Vitesse also benefited from outsourcing the manufacturing of CMOS products. What dragged down SDL was its expansion into the highly labor-intensive business of module assembly not an uncommon problem in photonics manufacturing. Photonics manufacturing tools Analogies between photonics manufacturing and the silicon industry can be helpful, but they can also be stretched too far. The most notable difference is that the photonics device sector is severely fragmented, with each market segment finite in size and thus unable to support its own specific enabling technologies. In contrast, as shown in , there are three unifying principles that underlie more than 90% of the worldwide semiconductor industry, namely: core "building blocks" in the form of transistors/gates; a single material system, silicon; and a dominant manufacturing process, CMOS. There are no analogous items for the photonics industry. In our view, this has served to dampen the growth of a monolithic photonics-manufacturing sector and has caused the photonics industry to "import" tools from other areas. There are four main types of "front end" processes for making the majority of photonics components: (1) semiconductors, encompassing both Si and III-Vs; (2) thin-film technology; (3) microelectromechanical systems (MEMS); and (4) micro-optics. Packaging and testing form the "back end". The tool set required for each of these technologies is obviously different from each other. We have compiled our own industry model, broken into four distinct segments: semiconductors, MEMS, thin film, and module assembly and test. We have chosen to ignore the software, materials, and micro-optics segments because they are too embryonic or fragmented to quantify at this time. While the total market for photonics manufacturing tools is expected to reach almost $4.5 billion in 2004, the market suffers from fragmentation, with the largest segment being automated assembly and test, which is projected to become a $2.6 billion market in a few years (see ). The second largest segment, semiconductor tools, will reach almost $1.5 billion in 2004, with a CAGR of 46%. The thin-film and MEMS tools segments are the fastest growing, but they make up a relatively small portion of the total market. Supply and demand The optical communications components and systems markets experienced phenomenal growth in the last few years, driven by the need of carriers to rapidly expand and upgrade their infrastructure to meet the anticipated explosive growth in bandwidth demand. Estimates for the annual growth rate vary, but 2533% is a useable benchmark. Meanwhile, the ASP of optical components is declining. Therefore, the unit growth of the optical component market is even more pronounced, potentially exceeding 50% per annum. Until late 2000, the optical component market had to live with rampant supply shortage. The backlog of optical systems was as long as six months in early 2000. Financially, the supply shortage has contributed to a unique distribution of profit margins among different participants in the supply chain. Photonics component companies seemed to enjoy significantly higher profit margins than their system OEM customers (see ). We believe that the shortage of photonics manufacturing capacity contributed to this disparity. Supply constraints What is the cause of these supply constraints? It is certainly true that demand growth has caught the industry by surprise. But once the situation was grasped, it proved surprisingly difficult to ramp up new production capacity for optical components surprising because given the rich stock valuation of the component companies and the eagerness of venture capitalists to open their wallets for op-tical start-ups, capital was not a limiting factor. Instead, we would blame the following factors:

Lack of scalability in the current manufacturing processes;

Shortage of technical talents for both engineering and manufacturing;

Absence of standardization throughout the supply chain.

Almost every photonics component has some manual processes in its production. Despite the high-technology nature of the products, manufacturing of optical components is most often compared to shoe making, in terms of labor intensity, lack of automation, and the lack of process optimization. Component assembly, fiber attachment (also called pig tailing), packaging, and testing are all notorious for their high labor intensities. Another culprit is variability. In mature industries like silicon, the variance in performance parameters among different parts is well understood and well managed. Typically, a system is designed with this variance in mind, and sufficient margin of error is built in to accommodate the idiosyncratic behavior of each part. In photonics, such system-level design practice is still immature, leaving the system performance dependent on the variability of each component. The variations are due in part to the manual labor involved in making the components, which in turn makes system-level assembly difficult to scale. The situation is further exacerbated by the difficulty of codifying knowledge in this field and the resultant dependence on artistic experience. Many photonics manufacturing processes require a certain art and experience that can only be obtained through practice and remain difficult to communicate. For example, an experienced coating engineer could easily increase the yield of thin-film coating by 100%, compared with a novice. Finally, the optical component market lacks standards. For example, the form-factors of thin-film dies vary among different vendors. This is widely recognized as one of the main inhibitors of manufacturing scalability throughout the supply chain, pertaining not only to the components, but also to the modules and systems. Some of this, in our view, stems from fundamental technological issues, while some of it is a reflection of the current economics in the early stage of the industry s maturity, which, we believe, can change rather dramatically over time. Three growth phases The good news, in our view, is that some of the aforementioned effects that have inhibited the growth of the photonics manufacturing tool industry will disappear in the near future. These transient effects include vertical integration, extreme secrecy and lack of standardization. In our opinion, these transient effects will diminish as the photonics industry goes through the current painful trough period. In other words, the trough of the photonics industry could be the genesis of the photonics manufacturing tool industry. We further hypothesize that the growth of the photonics industry has three phases: the brute-force phase, the synergistic phase, and the lockstep phase. The brute-force phase To cope with the supply shortage, photonics companies embarked upon what we will refer to as the "brute force" style of capacity expansion. This is characterized by rapid building of manufacturing plants, widespread hiring of low-cost manufacturing personnel, and purchasing of basic, non-integrated tools. Process engineering innovations, automation, and standardization are largely absent from this type of expansion since all of these factors have to take lower priority than time to capacity. Time to capacity is such an essential competitive factor in the supply-constrained industry setting, that we believe it may cause manufacturers to sidestep industrial engineering issues. To us, the brute-force phase of capacity expansion is rather unenlightening because it does not change the economics of the photonics industry, nor does it alter the industry structure from its current state. Specifically, it does not foster the development of an organic supply chain, including a sophisticated supplier base. In short, brute-force capacity expansion retards the maturation of the sector. If a company is able to expand its capacity with relatively minimal capital investment and sell almost everything it makes at an attractive margin, why should the company worry about increasing productivity? Demand and supply reach balance A major change in the photonics industry is on the horizon: we believe that supply and demand have come into balance sooner than many expected, especially given the recent downturn in the economy. For the first time in its young life, the photonics industry will have to cope with a supply glut. In the brute-force phase described above, manufacturing and process issues have held a relatively low priority, muting growth of the photonics infrastructure and providing opportunities only in isolated pockets. However, once supply catches up with demand (which is, after all, the goal of brute force capacity expansion) the industry s dynamics will be fundamen-tally altered as a demand-driven scenario emerges. In this environment, a successful company has to be either primarily a technology leader or a cost leader. In our view, it will be difficult to remain in the middle. The technology-driven company will likely continue to focus its resources on improving product performance and, to some extent, cannot afford to be distracted with manufacturing process issues. For commodity producers, on the other hand, cost and supply chain management become overwhelmingly important concerns because pricing will likely approach marginal cost. After the move to regions with low-cost labor, the only other fundamentally viable approach to cost reduction is through improvements in manufacturing efficiency. We believe that at this critical juncture we will see the emergence of a sophisticated, robust infrastructure in the photonics industry that is analogous to the tremendously sophisticated semiconductor equipment market that we see in the silicon industry. Since this infrastructure will play a critical role in enabling the future growth of the photonics market, we have chosen to describe it as "photonics enabling technologies", or PET. PET companies are those that supply equipment, materials, tools and general methodologies to the photonics industry. The synergistic growth phase We believe that the rewards associated with an outsourced technology supply base will quickly materialize and power another phase of rapid growth in the photonics industry, although this time the growth will be dramatically different from that of the brute-force period. This subsequent growth phase will be characterized by rapid changes in the photonics business model, including:

Rapid cost reduction, hence rapidly declining ASPs;

Improved manufacturing efficiency and scalability;

Reduced labor intensity combined with increased capital intensity.

As photonics companies outsource manufacturing technologies, they will gradually migrate toward higher value-added activities, such as:

Deeper, more proprietary technologies;

Innovative component design;

Module and system-level engineering.

In our view, the PET industry will experience extremely rapid growth around this time. Its growth rate will be the compounded effect of two drivers: first, the gradual but growing trend toward outsourcing, and second, the growth of the photonics market itself, empowered by the newly developed merchant supply of enabling technologies and the attendant benefits. Realizing that the two industries, photonics and its suppliers, will be mutually stimulating, we intend to label this part of the photonics and PET industries life cycles the "synergistic growth phase." The lockstep growth phase After the rapid growth of the synergistic phase, we believe that the industry structure and dynamics between photonics companies and the PET players will likely stabilize. Technologies will standardize, manufacturing platforms will consolidate, and the benefits of outsourcing will fully materialize. At this point, we would then expect the PET market to start tracking the photonics industry in terms of growth. We will label this phase the "lockstep growth phase." To get a feeling for how the industries will behave during this period of time and beyond, one only needs to study the current state of the semiconductor capital equipment industry. The market for semiconductor equipment has largely tracked the semiconductor end markets, with 22% of semiconductor revenue spent on capital spending year in and year out for more than a decade now. Conclusions Naturally, we would like to know exactly where we are in the industry s three-phase life cycle (summarized in ). We believe that the photonics industry has spent the majority of its short history in the brute-force mode, with only some embryonic activities in PET. Since the industry is transitioning from a supply-constrained situation to having more balanced demand and supply, it is natural to postulate that the PET market may be entering the synergistic growth phase, making this the high time for us to identify the investment opportunities in PET. At this point in time, the detailed segmentation analysis leads us to believe that investment opportunities in PET are largely limited to automated assembly and test, and specialized semiconductor process tools. While the thin-film and MEMS segments will also experience high growth, these sectors are too small to support independent investable companies. In the area of assembly and test, Newport, a merchant supplier of automated alignment and welding stations, is a potential investment vehicle. In the semiconductor segment, equipment companies focusing on compound semiconductors, such as Aixtron, Emcore and Riber, also offer investment opportunities. Other notable players include Thermo VG Semicon (part of a much larger company) and Applied Epi (currently evaluating an IPO), as well as materials suppliers such as AXT, IQE and Picogiga. An interesting aspect of the photonics manufacturing field is that most critical processes in use have been imported from other technology areas. Thus it is likely the major players that will emerge in the photonics infrastructure will have a heritage of serving other markets. Among other things, this means that there will be few "pure play" investments.