Editor’s note: Four years ago, Forbes ASAP published its first issue with a stunning prophecy by contributing editor George Gilder. Fiber optics, said George, had the potential to carry 25 trillion bits per second down a single strand. This represented a ten-thousandfold leap in carrying capacity over the 2.5 billion bits “barrier” long assumed by most experts in the field. What did George see that others had missed? One, a little-recognized (at the time) breakthrough called an erbium-doped amplifier, which keeps optical signals pure and strong over long distances. The other was a deep technical shift, with roots in the 1940s-era work of information theory pioneer Claude Shannon. If you believed Shannon, his logic dictated a new messaging scheme called wave division multiplexing. Though scorned by the experts four years ago, WDM now is emerging as the winner George had prophesied.
The real winners will be all of us, as the coming world of cheap, unlimited bandwidth unfolds and at last fulfills the true potential of the information age. Here is George with an update.
IMAGINE THAT IN 1975 YOU KNEW that Moore’s law, the Intel chairman’s projection of the doubling of the number of transistors on a microchip every 18 months would hold for the rest of your lifetime. What if you knew that these transistors would run cooler, faster, better, and cheaper as they got smaller and were crammed more closely together? Suppose you knew the law of the microcosm: that the cost-effectiveness of any number of “n” transistors on a single silicon sliver would rise by the square of the increase in “n.”
As an investor knowing this Moore’s law trajectory, you would have been able to predict and exploit a long series of developments: the emergence of the PC; its dominance over all other computer form factors; the success of companies making chips, disk drives, peripherals, and software for this machine. With a slight effort of intellect, you could have extended the insight and prophesied the digitization of watches, records (CDs), cellular phones, cameras, TVs, broadcast satellites, and other devices that can use miniaturized computer power. If you did not know precisely when each of these benisons would flourish, you would have known that each one was essentially inevitable. To calculate approximate dates, you had only to guess the product’s optimal price of popularization and then match its need for mips (millions of instructions per second) of computer power with the cost of those mips as defined by Moore’s law.
Merely by using this technique of Moore’s law matching and holding to it with unshakable conviction for nearly 20 years I became known as a “futurist.” Today I await the death of television, telephony, VCRs, and analog cameras with utter confidence as Moore’s law unfolds. You can tell me about the 98% penetration of TVs in American homes, the continuing popularity of couch-potato entertainments, the effectiveness of broadcast advertising, and the profound and unbridgeable chasm between the office appliance and the living-room tube. But I will pay no attention. Just you wait Jack Welch, Ted Turner, Rupert Murdoch, John Malone, and David Jennings the TV will die and you may be too late for the Net.
It is now 1997, and a stream of dramatic events certifies that another law, as powerful and fateful and inexorable as Moore’s, is gaining a similar sway over the future of technology. It is what I have termed the law of the telecosm.
Its physical base lies in the same quantum realm of eigenstates and band gaps that governs the performance of transistors and also makes photons leap and lase. But the telecosm reaches beyond components to systems, combining the science of the electromagnetic spectrum with Claude Shannon’s information theory. In essence, as frequencies rise and wavelengths drop, digital performance improves exponentially. Bandwidth rises, power usage sinks, antenna size shrinks, interference collapses, error rates plummet.
The law of the telecosm ordains that the total bandwidth of communications systems will triple every year for the next 25 years. As communicators move up-spectrum, they can use bandwidth as a substitute for power, memory, and switching. This results in far cheaper and more efficient systems. In 1996, the new fiber paradigm emerged in full force. Parallel communications in all-optical networks became the dominant source of new bandwidth in telecom. Like Moore’s law, the law of the telecosm will reshape the entire world of information technology. It defines the direction of technological advance, the vectors of growth, the sweet spots for finance.
AMERICA’S DARK SECRET FOR MORE THAN A DECADE, American companies have been laying optical fiber strands at a pace of some 4,000 miles a day, for a total of more than 25 million strand miles. Five years ago, the top 10% of U.S. homes and businesses were, on average, a thousand households away from a fiber node; now they are a hundred households away. ” However, the imperial advance of this technology conceals a dark secret, which has led to a pervasive underestimation of the long-term impact of photonics. Sixty percent of the fiber remains “dark” (unused for communications) and even the leading-edge “lit” fiber is being used at less than one ten-thousandth of its intrinsic capacity. This problem has prompted leaders in impact of fully the industry, from Bill Gates and Andy Grove to Bob Metcalfe and Mitch Kapor, to underrate drastically the impact of fiber optics.
Restricting the speed and cost-effectiveness of fiber has been an electronic bottleneck and a regulatory noose. In order for the signal to be amplified, regenerated, or switched, the light pulses had to be transformed into electronic pulses by optoelectronic converters. For all the talk of the speed of light, fiber-optic systems therefore could pass bits no faster than the switching speed of transistors, which tops out at a cycle time of between 2.5 and 10 gigahertz. Meanwhile, telecom companies could not deploy new low-cost fiber products any faster than the switching speed of politicians and regulators, which tops out roughly at a cycle time of between 2.5 years and a rate of evolution measurable only by means of carbon 14.
Nonetheless, the intrinsic capacity of every fiber line is not 2.5 gigahertz. Nor is it even 25 gigahertz, which is roughly the capacity of all the frequencies commonly used in the air, from AM radio to kA band satellite. The intrinsic capacity of every fiber thread, as thin as a human hair, is at the least one thousand times the capacity of what we call the “air.” One thread could carry all the calls in America on the peak moment of Mother’s Day. One fiber thread could carry 25 times more bits than last year’s average traffic load of all the world’s communications networks put together: an estimated terabit (trillion bits) a second.
Over the last five years, technological breakthroughs and legislative loopholes have begun to open up this immense capacity to possible use. Following concepts pioneered and patented by David Payne at the University of Southampton in England, a Bell Laboratories group led by Emmanuel Desurvire and Randy Giles developed a workable all-optical device. They showed that a short stretch of fiber doped with erbium, a rare earth mineral, and excited by a cheap laser diode can function as a powerful amplifier over fully 4,500 gigahertz of the 25,000 gigahertz span. Introduced by Pirelli of Italy and popularized by Ciena Corporation of Savage, Maryland, and by Lucent and Alcatel, today such photonic amplifiers are a practical reality. Put in packages between two and three cubic inches in size, the erbium-doped fiber amplifiers (EDFAs) fit anywhere in an optical network for enhancing signals without electronics.
This invention overcame the most fundamental disadvantage of optical networks compared to electronic networks. You can tap into an electronic network as often as desired without eroding the voltage signal. Although resistance and capacitance will leach away the current, there are no splitting losses in a voltage divider. Photonic signals, by contrast, suffer splitting losses every time they are tapped; they lose photons until eventually there are none left. The cheap and compact all-optical amplifier solves this problem. It is an invention comparable in importance to the integrated circuit.
Just as the integrated circuit made it possible to put an entire computer system on a single sliver of silicon, the all-optical amplifier makes it possible to put an entire system on a seamless seine of silicaglass. Unleashing the law of the telecosm, it makes possible a new global economy of bandwidth abundance.
Five years ago when I first celebrated the radical implications of erbium-doped amplifiers, skepticism reigned. I was summoned to Bellcore, where the first optical networks had been built and then abandoned, to learn the acute limits of the technology from Charles Brackett and his team. I had offered the vision of a broadband fiberspherea worldwide web of glass and lightwhere computer users could tune into favored frequencies as readily as radios tune into frequencies in the atmosphere today. But Brackett and other Bellcore experts told me that my basic assumption was false. It was no simpler, they said, to tune into one of scores of frequencies on a fiber than to select time slots in a time-division-multiplexed (TDM) bitstream.
Indeed, electronic switching technology was moving faster than optical technology. In the face of the momentum and installed base of electronic switching and multiplexing, the fibersphere with hundreds of tunable frequencies would remain a fantasy, like Ted Nelson’s Xanadu.
In 1997 the fantasy is coming true around the world. Xanadu has become the World Wide Web. The erbium-doped fiber amplifier is an explosively growing $250 million business. Electronic TDM seems to have topped out at 2.5 gigabits a second. TDM gear has suffered a series of delays and nagging defects and so far has failed in the market.
Electronic TDM failed not only because it pushed the envelope of electronics but also because it violated the new paradigm. In single-mode fiber, the two key impediments are nonlinearities in the glass and chromatic dispersion (the blurring of bit pulses because even in a single band different frequencies move at different speeds). Chromatic dispersion increases by the square of the bit rate, and the impact of nonlinearities rises with the power of the signal. High-powered, high-bit-rate TDM flunked both telecosm tests. By contrast, wavelength-division multiplexing (WDM) follows the laws of the telecosm; it succeeds by wasting bandwidth and stinting on power. WDM takes some 33% more bandwidth per bit than TDM, but it reduces power to combat nonlinearity and divides the bitstream into multiple frequencies in order to combat dispersion. Thus it can extend the distance or increase capacity by a factor of four or more today and can lay the foundations for the fibersphere tomorrow.
In 1996 the new fiber paradigm emerged in full force. Parallel communications in all-optical networks, long depicted as a broadband pipe dream, crushed all competitors and became the dominant source of new bandwidth in the world telecom network. The year began with a trifold explosion at the Conference on Optical Fiber Communication in San Jose when three companiesLucent Technologies’ Bell Labs, NTT Labs, and Fujitsuall announced terabit-per-second WDM transmissions down a single fiber. Sprint confirmed the significance of the laboratory breakthroughs by announcing deployment of Ciena’s MultiWave 1600 WDM system, so called because it can increase the capacity of a single fiber thread by 1,600%.
The revolution continues in 1997. At the beginning of January, NEC declared that by increasing the number of bits per hertz from one to three, it had raised the laboratory WDM record to three terabits per second. During 1996, MCI had increased the speed of its Internet backbone by a factor of 25, from 45 megabits a second to 1.2 gigabits. On January 6, Fred Briggs, chief engineering officer at MCI, announced that his company is in the process of installing new WDM equipment from Hitachi and Pirelli that increases the speed of its phone network backbone to 40 gigabits per second. Accelerating MCI’s previous plans by some two years, the new system will use a more limited form of wavelength-division multiplexing to put four 10-gigabit in-cause formation streams on a single fiber thread.
The first deployment will use existing facilities on a 275-mile route between Chicago and St. Louis, but the technology will be extended to the entire network. This move will consummate a nearly thousandfold upgrade of the MCI backbone, from 45 megabits per second to 40 gigabits, within some 36 months. Ciena, meanwhile, has announced technology that allows transmission of 100 gigabits per second.
Its February IPO was the most important since Netscape (market cap at the end of the first trading day: $3.4 billion). Why? Ciena is the industry leader in open standard WDM gear. During the first six months the MultiWave 1600 was available, through October 1996, the firm achieved $54.8 million in sales and $15 million in net income. (Lucent is believed to be the overall leader with more than $100 million of mostly proprietary AT&T systems.) At the same time, the trans-Pacific consortium announced that it would deploy 100-gigabit-per-second fiber in its new link between the United States and Asia. “The A powerful new player in these markets will be implications of Tellabs, currently the fastest-growing supplier the WDM of electronic digital cross-connect switches and paradigm go other optical switching gear. In a further coup, beyond simple following its purchase of broadband digital radio data pipes. The pioneer Steinbrecher, Tellabs has signed up all greatest impact 12 principals in IBM’s all-optical team. Headed of all-optical by Paul Green, recent chairman of the IEEE fiber Communications Society and author of the leading technology will text on fiber networks, and by Rajiv Ramaswami, likely come in coauthor of a new 1997 text on the subject, the consumer IBM group built the world’s first fully markets.” functioning all-optical networks (AONs), the Rainbow series. Tellabs now owns the 11 AON patents and 100 listed technology disclosures of the group.
The implications of the WDM paradigm go beyond simple data pipes. The greatest impact of all-optical technology will likely come in consumer markets. A portent is Artel Video Systems of Marlborough, Massachusetts, which recently introduced a fiber-based WDM system that can transmit 48 digital video channels, 288 CD-quality audio bitstreams, and 64 data channels on one fiber line. Aggregating contributions from a variety of content sourceseach on different fiber wavelengthsand delivering them to consumers who tune into favored frequencies on conventional cable, the Artel system represents a key step into the fibersphere. It can be used for new services by either cable TV companies or telcos.
The deeper significance of the Artel product, however, is its use of bandwidth as a replacement for transistors and switches. The Artel system works on dark fiber without compression. The video uses 200-megabit-per-second bitstreams (compare MPEG2 at 4 to 6 megabytes per second) that permit lossless transmissions suitable for medical imaging, and obviate dedicated processing of compression codes at the two ends.
A move to massively parallel communications analogous to the move to parallel computers, all-optical networks promise nearly boundless bandwidth in fiber. According to Ewart Lowe of British Telecom, whose labs at Martlesham Heath in Ipswich have been a fount of all-optical technology, the new paradigm will reduce the cost of transport by a factor of 10. For example, the optoelectronic amplifiers previously used in fiber networks entailed nine power-hungry bipolar microchips for each wavelength, rather than a simple loop of doped silica that covers scores of wavelengths.
As these systems move down through the network hierarchy, the growth of network bandwidth and cost-effectiveness will not only outpace Moore’s law, it will also excel the rise in bandwidth within computers—their internal “buses” connecting their microprocessors to memory and input-output.
While MCI and Sprint move to deploy technology that functions at 40 gigabits a second, current computers and workstations command buses that run at a rate of close to 1 gigabit a second. This change in the relationship between the bandwidth of networks and the bandwidth of computers will transform the architecture of information technology. As Robert Lucky of Bellcore puts it, “Perhaps we should transmit signals thousands of miles to avoid even the simplest processing function.”
Lucky implies that the law of the telecosm eclipses the law of the microcosm. Actually, the law of the microcosm makes distributed computers (smart terminals) more efficient regardless of the cost of linking them together. The law of the telecosm makes broadband networks more efficient regardless of how numerous and smart are the terminals. Working together, however, these two laws of wires and switches impel ever more widely distributed information systems, with processing and memory in the optimal locations.
“The all-optical WHAT SHOULD THE MAJOR PLAYERS DO NOW? network will triumph for the FOR THE TELEPHONE COMPANIES, the age of ever same reason smarter terminals mandates the emergence of ever that the dumber networks. Telephone companies may complain integrated of the large costs of the transformation of their circuit system, but they command capital budgets as large triumphed: It as the total revenues of the cable industry. is incomparably Telcos may recoil in horror at the idea of dark cheaper than fiber, but they command webs of the stuff 10 the times larger than any other industry. Dumb and competition.” dark networks may not fit the phone company self-image or advertising posture. But they promise larger markets than the current phone company plan to choke off their own future in the labyrinthine nets of an “intelligent switching fabric” always behind schedule and full of software bugs.
Telephone switches (now 80% software) are already too complex to keep pace with the efflorescence of the Internet. While computers become ever more lean and mean, turning to reduced instruction-set processors and Java stations, networks need to adopt reduced instruction-set architectures. The ultimate in dumb and dark is the fibersphere now incubating in their magnificent laboratories.
The entrepreneurial folk in the computer industry may view this wrenching phone company adjustment with some satisfaction. But computer firms must also adjust. Now addicted to the use of transistors to solve the problems of limited bandwidth, the computer industry must use transistors to exploit the nearly unlimited bandwidth. When home-based machines are optimized for manipulating high-resolution digital video at high speeds, they will necessarily command what are now called supercomputer powers. This will mean that the dominant computer technology will first emerge not in the office market but in the consumer market. The major challenge for the computer industry is to change its focus from a few hundred million offices already full of computer technology to a billion living rooms now nearly devoid of it.
Cable companies possess the advantage of already owning dumb networks based on the essentials of the all-optical model of broadcast and selectof customers seeking wavelengths or frequencies rather than switching circuits. Cable companies already provide all the programs to all the terminals and allow them to tune in to the desired messages. But the cable industry cannot become a full-service supplier of telecommunications unless the regulators give up their ridiculous two-wire dream in which everyone competes with cable and no one makes any money. Cash-poor and bandwidth-rich, cable companies need to collaborate with telcoswhich are cash-rich and bandwidth-poorin a joint effort to create broadband systems in their own regions.
In all eras, companies tend to prevail by maximizing the use of the cheapest resources. In the age of the fibersphere, they will use the huge intrinsic bandwidth of fiber, all 25,000 gigahertz or more, to simplify everything else. This means replacing nearly all the hundreds of billions of dollars’ worth of switches, bridges, routers, converters, codecs, compressors, error correctors, and other devices, together with the trillions of lines of software code, that pervade the intelligent switching fabric of both telephone and computer networks.
The makers of all this equipment will resist mightily. But there is no chance that the old regime can prevail by fighting cheap and simple optics with costly and complex electronics and software.
The all-optical network will triumph for the same reason that the integrated circuit triumphed: It is incomparably cheaper than the competition. Today, measured by the admittedly rough metric of mips per dollar, a personal computer is more than 2,000 times more cost-effective than a mainframe. Within 10 years, the all-optical network will be thousands of times more cost-effective than electronic networks. Just as the electron rules in computers, the photon will rule the waves of communication.
The preceding was adapted from the Gilder Technology Report, February 1997. For a newsletter subscription, call 1-800-888-9896.