Picking the winners

By Girish Mhatre

Forest fires are as periodic, inevitable and necessary in the man-made world as they are in the wild

Implicit in the choice of subject for this anniversary issue is our belief that the recent "forest fires" that have swept through the technology business may have been unique in their intensity but ultimately salutary in their cleansing effect. Thus, the ravages of the Internet and telecom-based economic collapses in the technology jungle will just as surely make room for new and sustained growth. The dense underbrush has been cleared, setting the stage for new technologies that will create new markets.

Our attempt in this issue is first to identify those enabling technologies that we think will have a disruptive impact over the next five to 10 years. What combination of features makes them better than incumbent technologies, and what do their feature sets portend when deployed in applications? Second, it is to assess their chances of doing so. Will they create pervasive new applications and new sources of economic growth or will they merely serve to excite briefly? The thrust of this article is an attempt to establish some set of conditions or qualifiers that can screen technology candidates for potentially disruptive impact. Our guide is history, which is rich with examples of technologies that did change the world and of technologies that did not live up to anticipations.

EE Times has observed and documented such phenomena several times over the span of its existence — the last 30 years. During that time the enabling technologies of silicon processing, of the microprocessor, of magnetic storage and of fiber-optic transmission, through natural, selective combination, contributed more than anything else to material wealth, at least in the industrialized world. In combination, they provided the tools — affordable tools, above all — for computing and communications. These were, all of them, "disruptive" technologies in that they upended the natural, evolutionary order of things.

And, though its invention predated the first issue of EE Times by more than 20 years, the transistor cannot be omitted in any discussion of disruptive technologies. It was the progenitor of all that we documented over the years and of much yet to come. In fact, the transistor serves as the splendid metaphor for all disruptive technologies. It replaced the active device of its time, the vacuum tube, by resetting all functional, performance and economic characteristics on a new curve.

The birth of all these technologies was accompanied by wracking economic pains not unlike the recent variety. There was always a forest fire; it was not until there was a shakeout that matched the abilities of certain players with the cold calculus of the market, that profitable and sustained growth could take hold. In every case, the higher the expectation, the steeper was the fall.

There were deadends and abject failures, of course. Certain technological advances, billed as world-beaters for their potential to spark discontinuities, simply did not make it. Bubble memories as storage devices and Josephson Junctions as switching devices come to mind. Others were simply ahead of their time and languished as laboratory curiosities until marketplace demand fortuitously intersected with their capabilities. Gallium Arsenide as a semiconducting material is an example. Its commercial potential lay unrealized for decades until high-speed communications demanded its performance. Sometimes an enabling technology did not make it on its perceived principal merits but ended up being perfectly serviceable — and commercially successful — in other applications that leveraged other features. Charge-coupled devices (CCDs) initially were developed for storage applications, but are now used almost exclusively in imaging.

At this point, it is important to make the distinction between enabling technologies and applications, as we see them. Generally, for the purposes of this issue, we consider applications as residing at a systems level — systems enabled by technologies at the component level. Thus, while we think that a post-PC personal appliance, one that combines the features of a computer, a voice phone and a data communications device in a portable, but wirelessly connected, package, is likely to have major economic — disruptive — impact, it's not the sort of technology we're talking about. That's an application by our definition. In contrast, the semiconductor, display and wireless technologies embedded into such an application are more likely to be the enablers of our focus.

So, how do we go about identifying potentially disruptive technologies? We can start by evaluating candidate technologies based on their merits relative to incumbents. But, one of the more curious things that becomes evident when examining the historical paths to deployment of new technologies is that they have often ended up spawning applications that were completely unforeseen. So, that approach in itself poses a conundrum: If the applications are unforeseen how do we know which technologies will enable them?

We can approach this by taking into account two trends: First, it is possible to identify — and extrapolate — certain "insatiable" demands in the context of current systems performance. Higher bandwidth and operational frequency, more memory, smaller size, lower cost, lower power consumption — the need to make systems "faster, smaller, cheaper" and energy efficient — are insatiable demands.

The other trend is the evolutionary path of current technologies extrapolated to their fundamental, physical limits. Or, to their economic limits; sometimes, it's just too expensive to tax a technology beyond a certain point. Silicon processing may be at there at this time. In any case, the question to ask is the following: Does current technology evolution keep up with the system-level demand curve for smaller, faster, cheaper, or does it diverge? If not, the resulting vacuum spurs the development and demand for a disruptive technology whose fundamental — or economic — limits are more distant. That technology, in turn, potentially creates newer — unforeseen — applications.

Today, in the insatiable-demand category fall new technologies that can address higher-bandwidth communications, either wired optical or wireless. We also believe that meeting the demand for bandwidth and mobility ultimately will create some hitherto unforeseen applications and services. In this category, as well, are technologies that enable discontinuous increases in storage density.

Another category of potentially disruptive technologies consists of those that can, for the first time, tap certain fundamental physical phenomena. In that category fall all the "small" technologies that explore little-known physical phenomena at atomic or molecular dimensions, or that shrink electromechanical systems to those dimensions. And here, though we may attempt to engineer these technologies into solutions for problems that exist, the more exciting possibility is that we'll solve problems that we don't even know can be solved.

But the potential for success of a new technology is not always based on what can be measured in scientific terms. History, again, demonstrates that the best technologies, on an objective basis, don't always make the biggest impact. Nowhere is this more evident than in the history of the internal combustion engine as the source of automotive power. The core engine technology, unchanged for much of a century, was never the best available technology at the dawn of the automotive age and has resisted replacement by objectively better technologies ever since. The electronics industry, too, has numerous such examples to contribute. In consumer electronics, as just one example, it is widely accepted that the Beta technology for video cassettes was superior to what became the standard — VHS. Similarly, the Intel microprocessor architecture at the heart of the PC was hardly the most efficient by a number of measures ("orthogonality," or lack of it, was a much-discussed concept in the early days of microprocessors) but emerged as the uncontested victor where it counted most — in the commercial arena.

So, while it might be tempting to hope that a better mousetrap always wins the day, the reality is that the better marketer and manufacturer has the edge. (It can be argued that the examples mentioned were triumphs of marketing and manufacturing over technological merits. They were, undoubtedly, but there's much more to it that was specific to the situations.) Oddly, this can cut both ways at the onset of a new technology generation. It turns out that, despite their expertise, the leaders in one generation of technology — the better manufacturers and marketers of their age — are at a significant disadvantage when it comes to extending their leadership to a new generation. In fact, it's hard to find a single instance of a leader in one generation of technology transitioning to a leader in the next. Simply put, the reason is that leaders in a particular technology generation have too much to protect. Culturally, therefore, they apply themselves to refining what they do best, not to upsetting the applecart.

Examples of this phenomenon are almost too numerous. The leaders in mainframe computing didn't lead in the minicomputers generation, whose leaders in turn gave way to others in the PC era. The leaders in germanium devices did not make it in the silicon age. The leaders in standard logic were not leaders in programmable logic. The list, of course, is endless. And that's an uncomfortable precedent for today's leaders. Whether Intel will maintain its dominance in the post-PC world is an open question, in that light. Will Dell, for that matter?

Thus, from the point of view of the technology challenger, the leaders of the current generation are vulnerable, but they do know how to defend themselves. Their defense is based on squeezing out even more advantage from their technologies by incremental, evolutionary advances and manufacturing efficiencies. Their weapon in the marketplace is lower cost to the user, which works only so far before gross margins become unsustainable. Consider the predicament that silicon processing is facing today.

As silicon processing reaches its limits (both physical and economic), various technology augmentations are being developed to extend its life, among them so-called low-k dielectrics, copper interconnects and silicon-on-insulator. (The latter among them is a variation of SOS, or silicon-on-sapphire, considered a potentially disruptive technology well over two decades ago.) It's important to note here that "augmentation" is almost by definition an evolutionary development, not disruptive. It adds cost, as does the industry's impending move to even larger wafer sizes, designed to increase manufacturing yields. But, in the process, the set up costs have ballooned. The price tab for IBM's latest fab runs well over a couple of billion dollars. Design and testing costs are rising as fast and contribute equally to the cost of chips. To maintain gross margins the leaders in silicon technology will have to find markets that are willing to pay for more expensive chips — high-end markets. That, quite possibly, would leave an opening for new technologies that make possible lower cost chips at the lower ends of the market.

What's not on the horizon, however, is a lower manufacturing-cost alternative to silicon. Instead, newer technologies will contribute to lowered overall costs — sometimes called replacement costs — by reducing the complexity and cost of design. That's the principal reason we've picked a third category of potentially disruptive technologies — those that rethink architectures and design approaches.

The operative word here is "replacement cost," which is the sum of the raw cost per function (per Megabyte of storage, for example) and infrastructure costs — cost of design, certainly, but also educational costs and others, even those that encompass social and regulatory issues.

Replacement costs decline with volume and that requires customers and markets.

Customers, however, aren't easily sold. One of the common underlying factors in the adoption of new technologies is that, regardless of an expanded set of benefits, customers see replacement cost as the primary determinant for initial deployment in systems. And, as incumbent technologies become even more competitive through evolutionary incrementalism, the harder it is for new technologies to break in.

So, it's a chicken-and-egg situation. New technologies have to find volume markets to achieve low replacement costs. But opening up those markets requires low costs to begin with.

All this demonstrates that the path to commercial success is at best convoluted and, at worst, unpredictable. One thing that's evident is that the odds are stacked against new technologies. There are always many more technology candidates in laboratories than commercial successes. There have always been more failures than successes.

The technology 'push' effect
Disruptive technologies often develop without clear markets that demand them. There are numerous examples of this, among them, more recently, the Web browser. Except for the people who thought it could be done, there was nobody else asking for it. Even the personal computer was a push application. Not until early spreadsheet and word processing programs became available (much later it was e-mail and Web browsing), was the PC something that anyone cared to ask for. Currently, nanotechnology and MEMS (microelectromechanical systems) potentially fall into the push category.

In contrast there are "pull" technologies. Pull technologies develop in support of well-defined applications and as replacements for incumbent technologies that may be reaching certain limits. The search for, and development of, such technologies intensifies when system-level performance demands (e.g. storage, bandwidth) can be extrapolated to exceed either the physical or the economic limits of incumbent enabling technologies.

The 'pull' of InP
Indium phosphide (InP) is a classic example of a "pull" technology. It's a semiconductor material that is likely to replace silicon as the basis of optical and electronic components needed to handle the higher data rates of tomorrow's optical network. InP is also uniquely qualified to integrate optical devices and electronic control circuits on one chip, inexpensively, thus enabling the creation of a new kind of mass-producible component, the photonic IC, a device that processes optical signals instead of electrical ones. Key to successful adoption of InP, however, is achieving lower production costs, a difficult task given the educational hurdles involved.

Ultrawideband technology (UWB) is a wireless technology that offers high bandwidth communications without spectrum constraints and is superior to conventional narrow-band, carrier-based technologies for short-range, high-data-speed transmission. And, it has the potential to be implemented with lower power dissipation and simpler designs than current approaches. UWB can also be used for precise measurement of distances or locations, making possible a whole new array of devices and services. Thus, UWB combines both the pull and the push aspects. The development of UWB is likely to be unique in that it is the only technology subject to government regulation.

The optical network starts with the light emitter or laser, which is the source of the optical signal that carries information along the optical fiber. We identify two developments in laser technology — tunable lasers and vertical cavity surface emitting lasers (VCSELs) — that will dramatically reduce costs and improve efficiencies in getting light into the fiber, and ultimately extending high-bandwidth access to "the last mile." Both these developments have been stalled to some extent because of the collapse of their customers. Both, however, are poised to rebound as the dust settles.

Nanostructures are squarely in the "push" category of technologies. It can be argued that as lithographically patterned semiconductor technology (so-called top- down manufacturing) approaches physical and economic limits, components constructed through atomically precise manufacturing methods (bottom-up manufacturing) could scale down much further. While individual devices (transistors and logic gates) have been demonstrated in the labs, recent advances have demonstrated more complex systems — made with nanowires, carbon nanotubes and even organic molecules — and better control over manufacturing processes. An unprecedented level of government and venture capital funding is speeding commercial potential. Meanwhile, intensive research is uncovering new materials and structures that exhibit unique new properties based on atomic phenomena. But the biggest problem with this technology is that hype may be overtaking reality, which usually sets the stage for unrealistic expectations.

MEMS (microelectromechanical systems) technology is already a big business in applications like accelerometers and disk drives. Now, as specialized manufacturing equipment becomes available, MEMS technology could become pervasive. The communications field, though temporarily stalled in terms of demand, is a major focus. But even beyond that, MEMS is poised to create radically new applications that combine electronics-based intelligence with acute sensing and precision physical control — all on one chip. MEMS technology uses the materials and processes of microelectronic fabrication to integrate extremely small mechanical components such as sensors and actuators with controlling electronics on a common substrate. Our enthusiasm for MEMS is based on the fact that manufacturing processes are not only standard, but do not stretch the state of the art.

Reconfigurable chip architectures deploy flexible hardware resources — resources that can be tuned or optimized to execute particular software protocols thus enabling products with greater computational efficiency and lower power consumption. They exhibit the potential for lower cost by using silicon real estate more efficiently.

Software defined radio (SDR) is a reconfigurable wireless communications technology that can accommodate diverse communications standards, again, "on the fly," for security (military applications) and convenience (consumer applications), and as a result, better manage spectrum resources. SDR-enabled user devices and network equipment can be dynamically programmed in software to reconfigure their characteristics for better performance, richer feature sets and advanced new services that provide choices to the end user.

System-on-chip (SoC) complexity — both in design and manufacturing — will drive the use of platform-based design methodology approaches, which are based on predefined abstraction levels and the reuse of precharacterized components, bus structures and architectures. A variety of styles are emerging, from purely programmable devices to architectures in which algorithms are mapped directly into specialized hardware blocks.

And, finally, mobile robotics: A brand new industry is about to be launched. It's hard to imagine now with robots still at the equivalent of the pre-Cambrian age but we think this industry could be could be as big as the worldwide automotive industry.