(4/23/97)

Paul Sagan has ridden hard over the wildest extremes of the Internet economy. In 1999 he stumbled into a $700 million fortune--on paper--by joining a brilliantly conceived startup called Akamai Technologies, which provided superfast digital delivery services to dozens of new Web sites. Then Sagan watched as the dot-com boom blew up. Akamai's Web site customers went bust, and Akamai plunged toward bankruptcy. Shareholders filed ten lawsuits, alleging a stock swindle. And then came the worst blow of all: Daniel Lewin, the brilliant young founder at the heart of the company, was killed by terrorists.

Akamai (from the Hawaiian word for "clever") seemed destined to become another casualty of the dot-com calamity. But six years later it has made an audacious comeback.

It powers the video wave now sweeping over the World Wide Web. Major League Baseball games, sitcoms from NBC, movie downloads from Starz and more zip across the Internet via 20,000 superfast servers that Akamai (nasdaq: AKAM - news - people ) has deployed in 71 countries. The company's video offerings are multiplying: Its servers "broadcast" live games for the National Hockey League in high definition.

Most of the big online sites on the planet use Akamai to deliver all sorts of digital goods faster than the regular Net can handle: Yahoo (nasdaq: YHOO - news - people ), iTunes, Microsoft (nasdaq: MSFT - news - people ), MySpace. Akamai parks data locally to cut microseconds from the travel time each packet takes to get to where it needs to be.

Hardly anyone notices. On a typical day one in five Web pages worldwide comes via an Akamai computer. The only clues for Web surfers that a page was akamade is the wait they didn't endure and the error messages that didn't occur. "The Internet is getting bigger and bigger, but it's not getting better and better," Sagan says. "That's our job."

Akamai's central role in speeding up the Internet has paid off: Revenue has doubled in two years, to $429 million last year; earnings are up 66% in the same period, to $57 million. Akamai stock, though distant from the $345 zenith it hit in the dot-com craze, is up almost a hundredfold from the bottom. It was at 56 cents in late 2002, and lately it has cleared the $50 mark.

Yet Paul Sagan, 48, says that what Akamai has done is nothing compared with what the company will pull off in the next few years. Akamai's chief since mid-2005, he believes this star-crossed company has arrived at a magical moment in tech history, able to at least double its capacity every year and in the process disrupt vast swathes of the global economy. Sagan believes Akamai has figured out superior ways to deliver anything digital: basketball games, phone calls, software, feature films, anything.

The giant incumbents in cable, telecom and direct satellite now push a triple play of services to consumers: TV, phone and Web access. To Akamai all those products are just different flavors of the same thing: digital packets. The Net will threaten these behemoths of the old guard.

"Anyone who relies on the fact that they own a scarce distribution network is going to face ten years of turmoil," Sagan says. If people want to get their TV shows over the Internet, the incumbents won't be able to stop them (Viacom (nyse: VIA - news - people )'s new legal assault on Google (nasdaq: GOOG - news - people )'s YouTube notwithstanding). He foresees an era of "emancipated consumers" who can get whatever content they want, whenever they choose. "They'll go right around the cable companies if they want." Sagan adds: "This was Danny's vision, and we are building it."

"Danny" is Daniel Lewin, Akamai's founder, an MIT-trained mathematician who conceived the algorithms and concepts that would form the business. On Sept. 11, 2001 Lewin, Akamai's chief technology officer, boarded the American Airlines (nyse: AMR - news - people ) flight from Boston to Los Angeles, on his way to a business meeting. Lewin settled into seat 9B in business class. One row ahead of him, on the other side of the aisle, sat Mohamed Atta.

A pair of hijackers got up from their seats in the second row and forced their way into the cockpit. They were followed by Atta, the only trained pilot in the group. No one knows for certain what happened next, but according to the 9/11 Commission, Lewin--a former commando in the Israeli army--likely tried to fight four hijackers singlehandedly, unaware of a fifth, seated directly behind him.

Flight attendant Betty Ong radioed the news of Lewin's battle to stunned co-workers on the ground: "Somebody's stabbed in business class." Half an hour later, at 8:46 a.m., the Boeing (nyse: BA - news - people ) 767 slammed into the North Tower of the World Trade Center. Daniel Lewin left a wife and two young sons. He was 31.

Paul Sagan, like the rest of Akamai's devastated employees, was left to question whether Danny Lewin's company had any chance of surviving his murder. Akamai had $101 million in cash, enough to last just over a year. Its stock, down 99%, was in danger of being delisted from Nasdaq. Lewin's vision of Akamai remaking the Internet suddenly seemed more like hubris than genius. "People told us what we were doing was both crazy and impossible," Sagan says. "It was a horribly dark period."

Yet through it all Sagan and others stubbornly clung to their belief in "the Big Idea"--Danny's idea.

By design, the Internet's power stems from its simple rules and communal nature. A network of networks, it uses a common language to let all "nodes"--computers of all kinds and designs and sizes--link to the whole, giving each one equal access. There is no central control or imposed hierarchy.

It is an abiding concept in communications, made famous by theorist Robert Metcalfe, that the value of a network increases with the square of the number of people hooked into it. That makes theoretical sense; the number of pairs you can make among N players is just about N ². The theory, though, doesn't take account of traffic jams. New Internet users add value to the network only if they can connect to one another, says Sagan. Akamai's big idea is that by rewriting the Internet's basic rules--making some computers smarter and more equal than others--it can let the Net grow infinitely large without breaking down.

Danny Lewin first happened upon the crazy idea of building a smarter Internet in 1996, just after he arrived at MIT. He was born in Denver and had grown up and gone to college in Israel. At MIT he was working with Frank T. (Tom) Leighton, a professor of mathematics who harbored a curiosity about how the Internet worked. Lewin, the mentally tough ex-commando, was hyperintense; his mentor, Leighton, laid-back and cordial.

Lewin and Leighton hashed out a basic idea: Connect computers to the far reaches of the Net, then program them to communicate with one another to spot better routes for getting e-mails, Web pages and other packets to where they needed to go.

Leighton, the head of the Algorithms Group at mit's Computer Science & Artificial Intelligence Laboratory, helped Lewin concoct ways to more efficiently deliver digital bits to their destinations. One obvious way was to make lots of copies of popular information and store them around the world. That way, when a user requested a popular picture, it could come from a computer across town instead of from one on the other side of the globe. It was an old idea--distributed storage was a feature of networks since the 1980s--but Lewin devised a way to do it far more efficiently.

Together Lewin and Leighton came up with lots of tricks to fix problems the designers of the original Internet never foresaw. They tackled problems like "flash crowds" (which can overwhelm computers storing a Web page or videoclip that suddenly is popular around the world) and the "fat file paradox": Why does data traveling at the speed of light often take such a long time to make it across the country?

Eventually they refined a business idea: a service that essentially would be the FedEx (nyse: FDX - news - people ) of the Internet. People could always trust the public Net to deliver their information cheaply. But others might be willing to pay Akamai a premium to deliver their content faster and more reliably.

The two men set up a "model Internet" in an MIT lab. They linked various computers and labeled them as different cities. To test their software they unplugged the wires connecting a pair of "cities" to simulate the traffic jams that occasionally bog down the Internet. Lewin tested his theories on the mini-network and tinkered continually with the formulas underlying it. He came to believe that their system could keep increasing the speed of the Internet forever.

Lewin, on a lark, parlayed their idea into a business plan and entered it in a contest at MIT's Sloan School of Management. They tied for fourth place. (Their prize: $100.) Then Lewin and Leighton got serious about building a real business, and in 1998 the duo at MIT attracted interest from a Boston venture firm, Battery Ventures.

The VCs, flummoxed by the mind-bending math behind the company, sent over an unpaid consultant to evaluate the academics' unusual idea: Enter Paul Sagan. He hit it off with the two computer scientists and proved just the right fit for the fledgling firm.

Lewin was prone to confrontation, often shouting to make his point. "Danny was the most competitive person I ever met," says Sagan. "He loved to yell and loved to argue." Sagan, by contrast, is a born manager and consensus builder. At age 11 he decided he wanted his family to take a vacation traveling cross-country and persuaded his parents to let him single-handedly design and organize the entire trip, his father, Bruce Sagan, recalls.

The young Paul reminded his father of another Sagan, Carl, the famous astronomer and Paul's second cousin, "If you think about Carl's ability to organize material and to see the variations and possibilities, you're describing Paul. He has a tremendous capacity to see down the line and understand where we are going," the father says.

Paul Sagan's first love had been journalism, not technology. His dad owns the Hyde Park Herald, a weekly in Chicago. Young Paul finagled a job working for a TV news department in high school; by 1987, at age 28, he was running news at the CBS flagship station in New York City. In 1991 Sagan jumped to Time Warner (nyse: TWX - news - people ), where he worked on some of the earliest Internet ventures. He helped launch Time Warner's Road Runner cable modem service, the first Web site for Time Inc.'s publications and the company's ill-fated effort to offer interactive TV in Orlando, Fla.

"Paul is the best manager I've ever worked with," says Walter Isaacson, Sagan's old boss at Time Warner. "And unlike a lot of CEOs, there's just not one ounce of jerk in him."

After Sagan was dispatched to check out Akamai in 1998, he encouraged Battery Ventures to invest; the firm put up $4 million for 17% of the startup. Then Sagan himself signed on, as employee number 15, becoming chief operating officer. George Conrades, an IBM veteran, joined as chief executive.

At the time Akamai's main customers were free-spending dot-coms that wanted Akamai to make their Web pages load more quickly. On the strength of those customers and Lewin's gauzy vision of building a smarter, infinitely expandable Internet, Akamai offered shares to the public in October 1999, in the heat of the dot-com frenzy. Akamai's investment bankers kept hiking the initial offering price, from $11 to $15 to $18 and finally $26, but it hardly mattered. The stock climbed 458% on its first day, giving it a market value of $13 billion. This for a company that had reported a grand total of $1.3 million in revenue. The stock doubled again in the next two months.

Lewin and Leighton became multibillionaires on paper. Sagan's 2.2% of the firm was worth $700 million. "We were arrogant," recalls sales chief Robert Hughes. "When you have one of the four best-performing IPOs in history, you get a little bit arrogant."

Then came the dot-com implosion. Customers with names like IshopHere.com and WinWin.com started going out of business. Conrades recalls wondering whether Akamai would have any customers left: "You would look at the customer list and say: 'What the hell are we doing?'"

With startups going broke by the hundreds, Sagan realized Danny's grand vision sounded to the world like all the other dot-com blather. At best Akamai looked like the folly of academic theorists who were not as clever as they thought. At worst it looked like a stock swindle built on a mountain of mathematical nonsense. The ten investor lawsuits accused the company of defrauding the public during its wildly profitable public offering, as well as of other misdeeds.

In the week following Lewin's death, Sagan spoke frequently with Conrades and Leighton; with Lewin, they always had run Akamai by committee, says Martin Coyne, a board member. "They were and are inseparable."

With Danny dead, they considered giving up. The trio began to ask one another hard questions. Their current customers were disappearing: Could they be replaced? After some debate each of the three agreed they could.

"Look around the room," Conrades said. "Do you believe we have good people?" That one was easy. Then Conrades asked the most important question: "Do you still believe in the Big Idea?" Danny's now ridiculed vision of building an infinitely powerful Internet--could they make it real? Conrades, Sagan and Leighton looked one another in the eye. Despite the trauma and turmoil that Akamai had endured, they still believed.

And there was this: "We had to do it for Danny," Sagan says.

Conrades and Sagan set plans to hunker down until bigger companies could replace the doomed dot-coms as customers and develop an addiction to the speed that Akamai provided. They laid off 800 of Akamai's 1,300 employees and cut their own pay to $20,000 a year. They made engineers do double duty as salesmen.

The company slashed operating costs from $2.6 billion in 2001 to $327 million in 2002. Even so, Akamai posted a loss from operations in 2002 of $182 million on revenue of $145 million.

By 2003 things began to turn around. Lots of real companies were interested in getting their data sent over the Internet faster and more reliably. The online speed needs of many older customers grew rapidly: At Apple (nasdaq: AAPL - news - people )'s iTunes music store, an Akamai client, users have downloaded 2 billion songs. Akamai started landing a wide range of new customers that could be counted on to pay their bills: the Canadian Broadcasting Corp., the U.S. Navy, Airbus.

Akamai's salesmen recrafted their pitch, telling established companies how much they could make with faster Web sites. Akamai started landing industrial companies who were selling more online. SKF Group, which sells ball bearings in Asia, used Akamai to speed the time their product's sales pages loaded, the better to entice impatient purchasing managers to use the site. In the next month SKF's Asian traffic jumped 30%, a number Akamai touts in its sales pitch to customers.

By 2005 the company was solidly back in the black, posting a profit of $14 million in the first quarter. Conrades took that as his cue to step down. The board asked Sagan take his place as chief executive. Sagan, ever humble, declined. "I thought there had to be somebody better," he says. "I spent a long time trying to convince them." Leighton, Conrades, Coyne and others insist Sagan's demurral was sincere. Says Leighton, Akamai's chief scientist: "It's not always the case that the best person for the job is the one who wants it most."

Sagan relented and took the job in April 2005, getting a raise to $400,000 for the year and a grant of 500,000 options at an exercise price of $13.33. (Current value: $18 million.) Sagan honed in on the most promising markets, where customers would be most likely to pay Akamai for delivering their bits.

Besides video, he made delivering software applications a top priority. The idea of storing software on the Internet is an old one. That way, software can be easily accessed from anywhere. But the concept has floundered on problems of reliability, security and speedy access. Sagan figures Akamai is perfectly positioned to solve all three problems, making a software application stored on Akamai servers act just like the application on a PC's own hard drive.

Under Sagan's leadership Akamai's invisible network has continued to encroach on more of the traditional, overtrafficked Internet. In 2006 its computers tripled the volume of data they delivered the year before. That growth is at least twice the rise in Internet traffic overall.

Akamai engineers in Cambridge, Mass. are constantly preparing newer, faster servers the size of pizza boxes--cheap off-the-shelf machines embedded with special Akamai DNA--and deploying them to the farthest edges of the network. Dozens of colleges in the U.S. have an Akamai server sitting on campus.

Every few months the algorithm writers in Cambridge inject better software into the global network to make it shrewder at picking routes for Internet traffic. Last year the software wizards added 100,000 lines of code to the 600,000 in each server. That upgrade doubled the network's ability to deliver bits.

Tom Leighton insists the Akamai grid can continue to double capacity, annually and easily--at the least. By adding more servers, making existing servers faster and upgrading software the company could boost capacity 500% a year or more if it needed to, he avers. Which is precisely what Daniel Lewin had in mind when Akamai began.

In the video age, too much is not enough. Akamai's global network now delivers 300 billion bits a second to computers around the world. That is enough to handle the constant barrage of Yahoo e-mails, iTunes downloads and YouTube-style videoclips. But now imagine that someone uses the Net to broadcast a TV show that gets a single Nielsen point (1% of the nation's 110 million TV households). The delivery would need a trillion bits, a terabit, per second. CSI would need 14 terabits. To deliver the same show in high def would require 100 terabits. That's 300 times Akamai's current output.

That mismatch will ensure the near-term future of giant incumbents in cable, telecom and direct satellite, with their dedicated networks expressly designed to carry video. But a decade out the Internet could catch up with a vengeance. Its power and smarts and bandwidth will continue to expand, and viewing habits will fragment further, eroding the notion of millions of people viewing a single program at the same time. The old delivery systems could be relegated to utilities, mindlessly ferrying data packets from one node to the next.

Ultimately, the biggest threat to Akamai may come not from the billion-dollar behemoths that its technology aims to replace but from tiny startups that seek to replace Akamai.

Peer-to-peer networks use the millions of PCs worldwide to create a significantly dumber but much larger version of Akamai's global distribution network. The creators of Skype, the Internet phone service that Ebay bought for $2.6 billion a year and a half ago, just announced a new P2P service called Joost that promises to deliver broadcast-quality video over the Internet, free of charge.

When one newcomer, MediaZone, started a Web site to send prime-time shows from China to Chinese expats living abroad, it signed on with Akamai. But it soon found that it could save more money by using peer-to-peer nets to deliver its bits. The more users it got, the more bandwidth it could save. "With 250 users, we use 85% less bandwidth, meaning we save 85% of the cost. With 25,000, we save 99%," says Graham Golder, the vice president of engineering at MediaZone ( FORBES, Mar. 26).

Other "content delivery networks," with names like LimeLight Networks, Savvis and Mirror Image Internet, have popped up. Many are cut-rate versions of Akamai, offering access to a much smaller global network in exchange for slightly less money.

Level 3, one of the biggest owners of fiber-optic cables, recently bought part of Savvis' network business, making it perhaps Akamai's closest rival. Level 3 is looking into combining Savvis' servers with the power of P2P networks similar to Joost. Many customers will prefer a hybrid network that steers traffic away from more expensive Akamai-style networks whenever possible, in Level 3's view.

Sagan points to the fact that Akamai has continued to grow faster than the Internet as a whole as proof that there will be bits aplenty to keep all these networks busy. He sees Akamai's global network as more reliable, smarter and more secure than P2P networks that make use of consumers' home PCs. "People aren't just looking to the cheapest bits. If they are, they aren't buying anything from me," says Sagan.

The vast array of companies delivering all sorts of bits has led to concerns in some quarters regarding the Net's ability to avoid a worldwide breakdown. "The unrelenting growth in Internet traffic during 2007 may overwhelm some of the Internet's backbones," according to a recent report by Deloitte Touche Tohmatsu. "The Internet is often regarded as an infinite resource. Unfortunately, this is not the case."

Oh yes it is, Paul Sagan counters. He says he often spends his weekend pondering whether Akamai has lived up to the goals Danny Lewin set for it. His conclusion: "Beyond a doubt we have proven that Danny was right. I sleep very well thinking that."

www.forbes.com/forbes/2007/0423/068.html