I (cont.) Basic Facts: The Component/Module Market
Primary characteristics of the components market:
 supply constrained
 high margins for component/module vendors
 barriers to entry high and getting higher
 time to market pressures
 labor intensive production with low throughput
 product qualification process difficult
 coopetition—with firms both customers and competitors
 demand consistently underestimated
RHK predicts that the optical component market should grow from $6.6b in 1999 to $23b in 2003, with component technology driving the optics market (DeutscheBanc Alex. Brown, Fiber Optics Review, 10/16/00). This year alone, the market for amplifiers and other widgets used to create terrestrial DWDM gear will grow 130% and no product area within the market will grow less than 80%. New technologies like array waveguides (AWG), with 1999 total sales of only $190m, RHK predicts will grow to $1.2b in 2003. (AWGs have the lowest price per port and the highest physical density of ports per unit—a good way to increase channel counts per system (RHK—2/00).
RHK sees tunable lasers and Raman gain modules growing from $280m to $960m by 2004, nearly 60% higher than original forecasts. Jay Liebowitz, RHK’s director of optical components says: “[We] forecast 80% growth for 2000. Everyone thought that too high; it turned out to be too low.” wwww.lightreading.com
Several factors contribute to the high barriers to entry.
 Trained engineers remain scarce. Mission-critical components must be Bellcore certified; this process is particularly arduous in the submarine market, and of the merchant vendors only JDSU and SDLI are qualified to sell into this market. The design, the product, and the factory must all be qualified and the process can take up to a year.
 Manufacturing is primitive, with throughputs in some products of less than 20%. Active components can be automated, but their packaging remains labor intensive.
 NG standards have not been set so components need to be tailored for individual systems vendors and even individual systems.
 High speed increases the complexity, making for a steep learning curve. As a result, profit margins of merchant vendors are equal to and sometimes higher (as in JDSU’s case) than those of their customers.
The players:
JDSU only makes components and modules, unlike Lucent and Nortel whose main business is optical systems. It is expected that NT and/or Lucent will spin off their “captive” components divisions.
Testing, Measurement
Systems Service
And Automation:
Newport
Veeco
Agilent
Melles Griot
JDSU (in-house)
Components:
JDSU
Corning
Lucent
Nortel
Subsystems:
JDSU
Avanex
Cyras
Systems Vendors:
Ciena
Corvis
Nortel
Lucent
Alcatel
Sycamore
Service Providers:
Qwest
AT&T
Verizon (etc.)
*adapted from The Fool (http://www.fool.com/Server/FoolPrint.asp?File=news/2000/jdsu001215.htm
JDSU’s products:
JDSU produces the basic building blocks for non-Sonet fiber-optic networks. Wavelength division multiplexing (WDM) allows multiple signals to be transmitted at slightly different wavelengths through a single fiber. It requires separate source lasers emitting slightly different wavelengths for each signal or “channel.” Each signal or “channel” carries a separate voice or data transmission. Once the signal is generated, modulators and optical amplifiers control and amplify the signal in the network to ensure that the transmission signal reaches its destination quickly and reliably. WDM systems have given way to DWDM, originally designed for 8 separate wavelengths or channels in 1996, DWDM are now being developed to carry over 160 separate channels—a statistic that demonstrates how young the field is and the rapidity of its technological development.
Until recently, JDSU has characterized, in financial reports, conference calls, and analysts meetings, its product lines by active/passive, with passive performing in the optical only realm and active in the optoelectronic. Generally, active components generate, encode, amplify or detect optical signals, while passive components are used to mix, filter, adjust and stabilize the optical signals. JDSU (with Etek acquisition) has 34% of passive market (couplers, taps, isolators, filters); with SDLI, 30% of pump module and about 80% of the 980nm diode lasers. Sixty four percent of the typical terrestrial edfa (erbium doped) dollar budget is spent on JDSU supplied components or modules (CIBC, 8/5/99).
Product Mix: percentage of revenue/operating margin/yoy% growth
Active: 29%/27%/113%
Passive:60%/43%/326%
Other:11%
(Bank of America)
The movement to packaging is key. A typical terrestrial high-powered (300mW) 980 chip may sell for around $500. A packaged module goes for $2,200 (CIBC). Subassemblies also facilitate compression of the product development cycle for JDSU’s customers, a point stressed at Supercomm2000 (Credit Suisse, 6/9/00).
More recently top management has moved to a characterization based on functions within the optical network. DeutscheBanc’s Raj Srikanth calls JDSU the key if not the leading player in all new technologies: tunable filters, tunable lasers, dispersion compensation, Raman amplifiers, and MEMs.
In CFO Tony Muller’s physics of photonics for economics and English majors he divides the module/component tasks into four sequential sets: transmitters, multiplexers, amplifiers, switches. Within each the level of technology (speed, reliability interoperability, remote control, and lack of conversion factor or OEO) results in lower costs and higher capacity (WitSoundView conference, 11/14/00).
Transmitters:
These modules combine source lasers, which power the initial signal, modulators, wavelength lockers and electronic drivers so that the signal is created and encoded in a single package. In addition to transmitters, JDSU also produces transceivers that are installed at the beginning and the end of the system. These modules combine transmitters with receivers so that signals can be generated and encoded or received and detected in a single package. The OC192 transceiver with 10Gbps is in deployment, and the OC768 with 40Gbps in development and testing. The physical effects that occur at high transmission speeds require a new level of components, like pulse generators.
Multiplexers:
DWDM thin film filters are the workhorses of DWDM. With the acquisition of OCLI, thin filters are scaling across the country. But the technology is broadening to include arrayed wave guide (AWG) and silicon based filters as well as interleavers where odd/even frequencies combine and the higher density translates to higher channel count within the available portion of the electromagnetic spectrum. As a result, costs can be amortized over more channels and more billable traffic.
Add-drop multiplexers allow systems to add and drop optical signals without reconversion to an electrical signal. For example, a system operating from San Francisco to New York can drop one signal in Chicago and add another, allowing for greater network flexibility. With Sonet rings, everyone gets off in Chicago.
Amplifiers:
Weak signals are amplified at optical speeds within the optical domain, amplifying all strands or wavelengths. As one goes from 16 channels to 160, amplification costs/wavelength decrease by 80%. JDSU supplies both Raman and edfa amplifiers. Raman source amplifiers use the Raman effect, inserting a high power laser signal into the transmission fiber so the fiber itself becomes a supplementary amplification medium. Raman pumps minimize crosstalk and eliminate active components. They also have the highest barriers to entry. Increased power requirements for Raman pumps make the learning curve steep as power and reliability are inversely correlated (CIBC, 8/5/99). JDSU/SDLI products have FIT (failure in time) scores of 100 years. When a 980nm pump laser in single pumped edfa fails, the entire system goes down. JDSU patented E2mirror passivation which completely suppresses COMD (catastrophic optical mirror damage) (CIBC).
Semiconductor amplifiers on chips complement erbium amplifiers. The first production line is ramping up in the Netherlands and customers are testing the products in a trial run. Eight years in development, these semiconductor amplifiers are 3-6 months away from shipping. Still “noisy,” they are suited to shorter distances of the metro market.
Switches:
Polymer, optomechanical, remote switching products are offered. MEMs, in Jozef Straus’ opinion, will be the platform for large-scale switching and bandwidth allocation on demand (CSFB conference, 11/29/00). JDSU took a lead in MEMs switching with its acquisition of Cronos. JDSU bought a MOT fab that manufactures the silicon device. Packaging, however, into a module is a challenge.
Customers:
LU, NT, and Alcatel each accounted for 10% of JDSU’s sales in the last quarter (11/26/00 quarterly earnings conference call). Nortel has roughly 40% of the total optical systems market, compared to about 20% for LU and 15% for Ciena. (Lu and NT are both customers and competitors.) Additional OEMs and established systems providers as customers include Cisco, Corning, Marconi, Motorola, Scientific Atlanta, Siemens, and Tyco (submarine). Emerging systems providers include Corvis, ONI Systems, Juniper Networks, and Sycamore.
The bulk of JDSU’s sales are into the US and Canadian markets with sales outside at 23% in 2000, 40% in 1999, and 38% in 1998, respectively.
Competitors:
JDSU confronts competitors on two fronts: competing against the clock and on the innovation front.
Lucent—competitive product: LambdaRouter, MEMS-based. LU’s optical-systems business, primarily Sonet-based, declined by 26% in the recent quarter when the market grew by more than 100%. LU missed the move to OC-192 (DeutscheBanc Alex. Brown), while its high capacity OC48 DWDM system was not ready to ship and customers went elsewhere quickly. Employees did too; LU’s turnover is near 20% with engineering talent exiting. moneycentral.msn.com
Corning: President John Loose expects sales at Corning’s components business to double to about $1billion this year (Bloomberg, 12/8/00), about 8% of total revenues. GLW lacks active component capacity to play at the module and subsystem level, which makes profitability difficult. Half of its component revenues come from edfa. GLW spent $3.6b on Pirelli’s optical components unit—a business with one customer, no profit, and less than $25m in annual sales and there are rumors that Pirelli is not exactly running like fine-tuned Italian sports car (http://www.lightreading.com/document.asp?doc_id=1283). GLW is just about alone in the ability to bundle fiber and components together (DeutscheBanc, 10/16/00)). GLW has 50% of the fiber market (LU has 20%), and its LEAF (large-effective-area fiber), although three times more expensive, cuts network costs by 30%. 1,700 PhDs at Sullivan Park
Avanex. Epoch Partners projects revenues close to $500mm by 2003. Just now moving products like PowerShaper (dispersion compensation) from pilot production into commercial volumes. Suite of products built around filtering technology: PowerFilter(dielectric based filter). Second commercial product is PowerMux, a NG wavelength division multiplexer with 800 channels at 10Gbps. Main customer for PowerMux is MCI Wcom, but AVNX is now shipping the product to Nortel, Fujitsu, and Sycamore and in testing with LU, Siemens, and Cisco. PowerMux uses Fabry-Perot interferometer technology so that frequencies are self-aligning. Self-aligned silicon gates were critical to the development of the integrated circuit at Intel and self-alignment will be equally critical in intelligent optical networking. PowerExchanger, a “switchless” optical add-drop multiplexer based on PowerMux design is tracking ahead of plan. PowerRelay is an optical amplifier for long-haul applications (http://www.epoch.com).
Cyras—optical switching systems for metro market. Closely held, to be acquired by Ciena.
Bookham—innovative production techniques. Throughput for optical components is around 20%, but Bookham integrates the functions on a single silicon chip in a process that is faster and cheaper, but also yields about 80%. Bookham licensed the technology to NT and LU in 1997 and others since (redherring, 12/18/00). |
