(FORTUNE Magazine) – "Bandwidth is kind of like a drug," says Dana Waldman. "The more you get, the more you want." As CEO and founder of a private communications startup that's building a potent new way to deliver this drug, Waldman knows just how addictive the stuff can be. He also knows that the bigger the dose, the better.

Over the past few years, bandwidth--the capacity for carrying digital information--has been doubling every nine months, roughly twice the growth rate of computing power. San Francisco market research firm RHK estimates that the total U.S. traffic carried over the Internet will increase from about 350,000 terabytes a month last year to 16 million by 2003--enough to hold the contents of the Library of Congress more than 100,000 times over.

When it comes to satisfying the national jones for broadband, nothing works like the fiber-optic strand: not copper, not cable, not high-frequency radio. For pure speed or data-carrying ability, nothing is as good as glass. As with most rapidly evolving technologies, optical networking's hottest innovations are found not in the blue-chip companies at the center of the economy but in the nimble, inspired startups around its edges. Giants such as Cisco Systems, Lucent Technologies, and Nortel Networks have no choice but to supplement their internal R&D efforts by buying innovation on the open market. Thus, startups that develop a way to boost bandwidth often sell for billions of dollars--even before they've moved a single product. Those that manage to stay independent and go public fetch much more: At press time Ciena, Corvis, and Sycamore Networks, for instance, each boasted a market capitalization of about $30 billion.

Waldman's private startup, Centerpoint Broadband Technologies, hopes to be in that league one day. Centerpoint is building gear that can increase traffic in fiber-optic networks as much as eightfold, at about half the cost of competing technologies. Numbers like those have drawn some very smart money to Centerpoint, including well-respected Silicon Valley venture capitalist Cliff Higgerson of ComVentures, who was a first-round investor in Ciena six years ago. Bill Berkman, the founder of Teligent, is backing Centerpoint too, as are Cisco Systems and SBC Communications. This is the type of company that makes IPO investors drool.

Birth of a Fat Idea

The genius of Centerpoint comes down to a flat gray rectangle about the size of a pizza box; to really understand the scale of Waldman's inspiration, you have to look inside it.

The bandwidth of an optical network is determined not only by the speed of light traveling through glass but also by how much light the fiber can accommodate at any given nanosecond. Because the speed of light is constant--and because fiber strands are most efficient when superthin--efforts to improve bandwidth have centered on finding ways to push more information through the same tiny pipe.

Back in the early 1990s, fiber could carry only one wavelength, or color, of light at a time. Data were packed into the fiber as digital pulses created by a laser turned quickly on and off, the optical equivalent of ones and zeros. This allowed some 2.5 billion bits of data per second to pass through the fiber. Over the past decade, however, companies such as Ciena, Lucent Technologies, and Nortel Networks developed a way to send multiple wavelengths down the same fiber--a technique called wavelength division multiplexing (WDM). Initially, WDM could split the light going down a fiber into only two colors, but it quickly became four, then eight, 16, 32, 64. There were so many colors that it began to be referred to as dense wavelength division multiplexing (DWDM). Today the most advanced DWDM system can carve light into as many as 160 colors. Since the wavelengths do not interfere with one another, each can carry as much data as an undivided, single-color fiber strand.

Now take a step back and look at a state-of-the-art optical network as a whole: Thousands of digital signals emanating from homes, office buildings, Internet service providers, and phone companies travel at varying speeds (some are light, some are electrical pulses) until they reach what's known as a TDM (time division multiplexing) box. There they are combined and converted into optical pulses and then sent through a single-color fiber-optic strand. The strands from multiple TDMs (technically up to 160) are then brought together at a DWDM box, where the information carried by each is converted into a different color. The DWDM box then ships the collection of colors down different lanes along a single piece of fiber, essentially creating a corresponding number of lanes where once there had been only one. At the receiving end of the transmission, the signals are broken down into their constituent parts and directed to their destinations.

All of this is, frankly, pretty mind-blowing already. Now along comes Centerpoint with a way to cram multiple frequencies into each color of light. This is known as frequency-division multiplexing (FDM), and it represents a completely new way of increasing the bandwidth of fiber-optic networks.

Centerpoint's FDM equipment--the pizza box--is called Lens; it sits in the nodes between strands of fiber, between the TDM and the DWDM gear. The Lens box adds one more step to the aggregation of light beams (or to the splitting of them, depending on whether it is on the transmitting or the receiving end of the signals). It converts light beams from the TDM boxes into radio frequencies and uses sophisticated modulation techniques to pack the frequencies closely together by slightly staggering them for the billionth of a second or so that they're within its circuitry. Then the frequencies are "stacked," essentially creating multiple lanes of information within one composite radio wave. The Lens box then converts the radio wave into a continuous analog wave of light (as opposed to the digital pulses that it received); when that analog wave, with its many lanes of data, enters the DWDM box downstream, the DWDM box shuttles it along one of the color lanes in the outgoing fiber. Each color coming out of the DWDM box now carries up to eight times the information it could carry without FDM technology.

Centerpoint is not the only FDM dealer on the block. Another company, Kestrel Solutions, has raised even more money--about $312 million to date, vs. $57 million for Centerpoint--and is closer to getting its product to market. But Kestrel's box isn't as fast as Centerpoint's and the rival company doesn't own its modem technology. So as FDM equipment advances to faster and faster speeds, Kestrel will have to rely on other modem suppliers to keep up with Centerpoint.

But both companies could do fine. The potential market is vast, and both flavors of FDM--Centerpoint's and Kestrel's--have some very tangible advantages over competing technologies. For one thing, counterintuitive as it may sound, the conversion from digital to analog light in FDM technology has made it easier to create bandwidth. When traditional digital optical gear gets up to speeds in the billions of bits per second, it becomes exceedingly difficult (read: expensive) to turn on and off the lasers that generate the pulses of data. FDM's analog system sidesteps that problem by using modulation--shades of gray, if you will--in place of hard pulses.

Also, FDM can be used to either accelerate a DWDM system or--because it can increase the capacity of a single-color strand of fiber--replace it. Especially in new metropolitan fiber rings, one of the fastest-growing segments of the optical networking market, FDM can actually be a cheaper alternative to the very pricey DWDM boxes and other required equipment.

Centerpoint's technology can thus speed up existing networks and extend their useful lives--more cheaply. That speaks to one of the central problems of optical networking in a country craving bandwidth: "Carriers have increasing transport needs that are growing exponentially, but their revenues are only growing linearly," says Centerpoint's vice president of marketing, Roselyne Genin. "They have to leverage their optical expenditures." Nothing like being in the right place at the right time.

A Fat Idea Gets Traction

"I think of us as a turbocharger," says Waldman. "We turbo-charge the Internet." He is sitting in his San Jose headquarters on an ultramodern blue plush chair with steel wheels. In keeping with the work-is-play decor, his bookshelf holds Silicon Boys and The Gorilla Game, as well as a collection of limericks, a gumball machine, and a box of Corn Pops.

Such frisky trappings notwithstanding, Centerpoint traces its origins to the drab halls of Loral (subsequently swallowed by Lockheed Martin) and the no-nonsense world of military satellites. In 1993, Waldman headed up a classified government-funded project to develop a high-bandwidth satellite modem that could transmit data at about ten billion bits per second through a very narrow slice of radio spectrum. Four years and $100 million later, Waldman and his team of engineers had a working satellite modem, and Waldman was rewarded with the reins of a business unit with more than $80 million in revenues and 400 employees that deployed and managed satellite networks. He was 33 at the time and one of the youngest executives in all of Lockheed.

The trick to making high-speed modems for satellites is filtering out the real signals from all the noise. If you can do that across the void of space, then packing a bunch of relatively clean signals down an optic fiber and sorting them at the other end is a piece of cake. At Centerpoint, Waldman has adapted the same modulation techniques he developed at Lockheed to create gear that can transmit and receive data traveling at an aggregate 20 billion bits per second.

Waldman's idea was too good to be denied. Even Lockheed, whose record with spinoffs was abysmal, saw the opportunity in its lap. Waldman had known all along that his technology had the potential for massive commercial appeal, because the algorithms and signal processing could easily be applied to other communications fields. In May 1998 he urged Lockheed executives to let him come up with a business case to capitalize on the technology. It worked: He got the go-ahead and enlisted the help of an incubator in Florida called Military Commercial Technologies (Milcom) that specializes in spawning startups from the defense industry.

Waldman and Milcom explored different options. The satellite market was the most obvious, but it had been slow to take off in the face of rapidly evolving and expanding terrestrial networks. Cable modems were dismissed out of hand, given that they were already beginning to appear on the market. That left two choices: broadband fixed wireless modems and optical networking gear. (Fixed wireless technology is a way to connect businesses and homes to broadband hubs via radio waves; optical networks, of course, make up the backbone of the Internet.) Demand for broadband was soaring, so the markets for fixed wireless and optical networks were likewise ready to lift off. Centerpoint pursued both markets, but optical became its primary focus.

Just as Waldman and the folks at Milcom were thinking this through, Lockheed decided to embark on a massive reorganization, which proved a fortunate catalyst for the project. The business Waldman worked for was going to be subsumed into a bigger entity; as a rising star, Waldman was one of three candidates being considered to run one of the new divisions. He and the other two candidates received a formal memo summoning them to a meeting with one of Lockheed's honchos.

More concerned with the fate of his project than with a promotion, Waldman spent most of the days leading up to the meeting trying to figure out how to structure a deal that would turn his technology into a business. The only way it could work, he believed, was to create an entirely new company. He sketched out his plan on a piece of paper, writing "New co." in the center, flanked by "Lockheed" and "VCs," with arrows connecting the various entities. The idea was that in return for equity, Lockheed would pump in its technology, and venture capitalists would pump in money.

On the day of the meeting Waldman found himself sharing an elevator with the sector president, Art Johnson, who remembered Waldman's presentation from a few months before. Johnson offhandedly asked about the project: "What do you think you want to do with that?" Waldman produced the crinkled sketch from his navy blazer and said, "This is what I want to do with it." Johnson looked it over and nodded.

A couple of hours later, his confidence boosted by the elevator encounter, Waldman showed up for his audition with one of Johnson's deputies. When the executive conducting the interviews asked for the bundle of paperwork he'd been instructed to bring, Waldman said he hadn't filled it out.

"Where's your package? Didn't you get the memo?" asked the executive incredulously.

"Yes," said Waldman, "but I don't want to talk about my package." Once again he drew the paper from his pocket and proceeded to explain his spinoff. The executive, still perplexed, asked, "Don't you want to be a candidate to run this?"

"No, I want to do this instead," insisted Waldman. His refusal to even try for the plum position was unusual (if not alien) in the defense industry. But that 20-minute meeting stretched to an hour and a half. At the end, Waldman was told that there was no guarantee the deal would happen. "I'll take my chances," he replied. On his way out, he passed the third candidate, furiously cooling his heels.

His employer's state of flux played to Waldman's advantage. He eventually got approval to spin off what became Centerpoint, with Lockheed agreeing to license its technology in return for royalty payments and a small equity stake. Lockheed also incubated the company in its San Jose offices and for the first few months paid the salaries of the five engineers Waldman took with him.

The prospect of a new technology that could help alleviate the bandwidth problem was even more appealing to the investing community outside the military-industrial complex. In the summer of 1999, as a favor to Milcom, Waldman accompanied some of its partners to a meeting with U.S. Trust in New York. Milcom wanted to get U.S. Trust to invest and was trotting out Waldman as one of its showcase CEOs. The bankers were so impressed with his pitch that the next week one of them visited Centerpoint, which was preparing for a second round of financing. U.S. Trust generally doesn't invest directly in companies, but it does invest in several venture capital funds. One of them is ComVentures, whose partner Cliff Higgerson was ranked as one of the top wealth-creating VCs in the Valley for 1998 because of his investments in Ciena and other communications equipment companies. The U.S. Trust banker promised to broker an introduction.

A few days later Higgerson called, and the two met for breakfast at Il Fornaio, a VC hangout in Palo Alto. Within venture circles, the soft-spoken Higgerson is a legend. Besides Ciena, he had also backed AOL, MCI, and Tellabs. Getting Higgerson onboard would be a real coup for Waldman. Forgoing his charts and cinnamon toast, he launched into his pitch ad-lib as Higgerson ate his oatmeal. When he finished, he tried to act tough, telling the VC that he'd need a quick decision because he had other investors lined up.

Higgerson looked up mildly from his oatmeal and asked, "Would this afternoon be okay?" That afternoon Higgerson closed the deal on the phone, agreeing to invest $5 million. (He subsequently invested another $6 million.) "Being a first-round investor in Ciena," he says, "I'd say there is a similar amount of opportunity here."

After Higgerson, the deluge began. Cisco Systems also invested in that second round, as did Bill Berkman of the Associated Group, who agreed to join Centerpoint's board. In March, Waldman raised another $40 million, adding Menlo Ventures and U.S. Venture Partners as investors. Last June, SBC Communications began evaluating Centerpoint's Lens product and also became a strategic investor.

"Everyone's throwing money at us," crows Waldman today. He officially unveiled Lens in August, and plans to begin shipments by the first quarter of next year. An IPO could follow soon after--if Centerpoint doesn't get acquired first. Either way, the technology is sure to keep feeding our bandwidth demons. As Waldman knows well, once we get a little taste of the good stuff, there's no going back.

FEEDBACK: erick@ecompany.com