Capex vs Opex
But perhaps the component vendors are barking up the wrong tree? I’d like to put forward the case for paying more attention to lowering capex than opex.
The drive to lower opex in optical networking originated from analyses that Ciena Corp. (Nasdaq: CIEN) conducted in the late 1990s on the predicaments of U.S. interexchange carriers. Ciena showed that data traffic was rising rapidly, revenues from data traffic were growing at one-seventh this rate, and costs associated with data traffic were growing faster than revenues.
McKinsey & Co. and Goldman Sachs & Co. continued this work in 2001, when they maintained that service providers could not solve their financial problems simply by reducing capex. Service provider costs needed to fall at 30 percent a year, but since capex made up less than 30 percent of total spending, service providers needed to reduce opex as well. McKinsey and Goldman Sachs concluded that service providers needed to reduce operating expense at a rate of 24 percent a year.
Equipment manufacturers have embraced these recommendations and have been promoting their systems as lowering capex and opex. One optical networking equipment manufacturer – Photuris Inc. – stated at NFOEC 2002 that its new product could save 50 percent on opex versus systems in use today. In addition, the manufacturer maintained that for every dollar invested in its equipment, the product saved as much as seven dollars in opex.
These assertions are encouraging for the optical networking industry. They also raise expectations that optical components can be developed that have parameters that create quantifiable opex savings. Defining such parameters, however, seems elusive. In addition, statements one hears at different levels of the optical networking supply chain suggest that searching for opex parameters may not be as rewarding as concentrating instead on those that relate directly to capex.
For example, service providers are intensifying their payback requirements on new equipment. New equipment can often incur increased initial costs, for features that enable greater savings during the life of the equipment. If a service provider had required a payback period of three years on a new kind of optical networking equipment, then the provider may now be requiring payback to be demonstrated in just one year. The shorter recovery period places more of a premium on delivering the biggest financial advantage upfront – in other words, cutting capex.
Moreover, it is not clear how well service providers can quantify optical networking opex in legacy networks or how much these costs can really be lowered. Technology planning engineers at service providers point out that opex incorporates many factors beyond the direct cost of provisioning, maintaining, or protecting a wavelength. These engineers agree that equipment that allows for point-and-click lighting of wavelengths does have the potential to provision services faster and minimize the cost of sending technicians into the field. Service provisioning costs, however, are complex and incorporate many other factors. Lighting a wavelength is just one of many costs in the process, with others from operations, administration, provisioning, and maintenance creating bottlenecks that are more significant to the total cost. Technician costs are also not variable, as service providers cannot readily eliminate this function. Union contracts with service providers, as well as prudence against unforeseen scenarios, encourage service providers to retain their technicians.
Technology planning engineers also point out that that the remote, automated capability of next-generation optical networking equipment can increase service provider costs. Using this feature requires that DWDM equipment and routers be populated with transceiver boards in advance of live traffic. Yet, having a DWDM transmitter and receiver, for example, sitting idle until traffic is turned up is asking a service provider to leave tens of thousands of dollars of inventory unused. In a tight economic climate, a service provider is motivated to preserve cash rather than purchase capital that is not immediately generating revenue.
Such complexities concerning opex suggest that the optical component vendor should concentrate on what can be quantified in selling to the intermediary, the equipment vendor. There are clearly some parameters that do affect opex that can be quantified. Size and power consumption are among two of the obvious ones. Real estate and cooling costs are significant expenses for service providers, so any savings that component manufacturers can provide will help everyone.
Optical components that are inherently geared for lowering opex by being dynamically and remotely tunable, however, are struggling to take off. One can read about design-ins of widely tunable lasers and dynamic spectrum equalizers, for example, but their vendors complain that revenues are disappointing. Worse, they are not taking market share from earlier, less sophisticated versions of these products.
An executive from a tunable laser manufacturer told me privately at NFOEC that his full C-band products were simply overkill for any business of significant revenue. He said that six nanometers was sufficient for applications that could actually ship, as these applications were for sparing and inventory management. The incremental cost for eight 100GHz tuning range was sufficiently small to justify these static applications, but the price premium for an even wider tuning range was just too high. The slow acceptance of optical components specifically for highly automated optical networks suggests that less dynamic optical components provide a more valuable mix of capex and opex savings.
I recommend to my optical component clients that they rely on the mix that has worked most successfully over the history of optical communications. A reduction in the cost per bit per second per kilometer is the argument that justified single-channel TDM transmission in the 1980s and WDM in the 1990s. Architectures have varied and are still evolving today, but the metric of $/bit/s/km continues to be the most persuasive, largely because it is the most simple to quantify and use for comparisons.
The optical component vendor can differentiate himself in $/bit/s/km by improving either the numerator or the denominator. The vendor can lower the price, while keeping the same distance-bandwidth product; or the vendor can preserve the price, while adding dBs to the link budget. Of course, with pressure on saving money so strong these days, customers value improvements to the numerator more highly.
How much of an improvement do equipment manufacturers require to be motivated to adopt a more powerful technology or more economical product? During the boom, one heard from the service provider community expectations of 10x improvements in first-installed cost. These targets were, however, for complete overhauls of existing networks rather than for incremental improvements. No service provider today has the available capital to invest in a brand new network.
Current targets seem to be more consistent with historical expectations, based on incremental buildouts, upgrades, and rehabilitations. Optical component vendors, prior to the boom, were used to equipment vendors expecting reductions of at least 15 to 20 percent to justify engineering redesigns. Today, with competition intense for every available socket, equipment manufacturers are expecting much steeper declines. Some equipment manufacturers are saying that the overall material cost of the board has to fall at least 15 percent to justify a redesign. If the optics on the board makes up just 25 percent of the material cost of the board, as may be the case in a metro system, then the optics have to fall 60 percent so that the net cost declines 15 percent.
The challenge to the optical component vendor is to meet these stiff expectations in a differentiated, sustainable manner. Significant improvements in cost per Gbit/s per km are difficult to achieve, especially in the numerator. Still, offering improvement on the price-performance ratio is more valuable than making promises that are not as easily quantifiable. Any optical component sales engineer will tell you that in a competitive bid, link budget and price virtually always trump footprint and power consumption. Specifications even more remotely removed from first cost, such as those relating to maintenance, find themselves even lower on the priority list. After achieving leadership on price-performance, the vendor also needs to offer other benefits that demonstrate its product as distinctive in a crowded field. In addition, the vendor has to prove long-term viability, either with cash reserves or a partnering relationship with a major supplier. These three areas are essential for success.
In conclusion, optical component manufacturers should definitely address opex issues in their development and manufacturing of new products. Increasing functionality, reducing size and power consumption, decreasing the number of fiber connections, enhancing reliability, etc., are all valuable to the health of the industry. Given the questions raised above, however, the component manufacturer seems more likely to increase his ability to increase sales by concentrating first on lowering capex and making decisions concerning opex subordinate. Such recommendations on decision-making criteria may seem to place the optical component vendor’s health ahead of those of the industry, but the ambiguities in transforming service provider needs into quantifiable parameters at the optical component level strongly suggest such behavior.
— Jay Liebowitz is founder and president of Liebowitz Strategies, which helps companies with strategy, business development, and positioning for increasing revenue and achieving corporate financing milestones. Liebowitz – who can be reached at [email protected] – has many years of experience providing advisory services, consulting, and market research services in optical and electronic components and subsystems, and has also managed businesses at Lasertron (now Corning) and Epitaxx (now JDS Uniphase).