(FORTUNE Magazine) – Wearing a cordless, ear-attached mike that makes him look like a rock star in a suit, 54-year-old John Chambers jumps onto the stage with the energy of a man half his age. Here, in front of some 250 customers, investors, and media people, the CEO of Cisco Systems is introducing his company's latest and greatest product, and you can tell from his speech that, for him, it is a very big day. For about 30 minutes he ranges around the stage and into the audience, filling the air with superlatives. He tells the crowd how the CRS-1--that's the product, which looks like three refrigerators side by side--will "enable the next generation of Internet usage"; how it leaves Cisco's competitors in the dust; and how it represents just the tip of the iceberg in terms of products and innovations from Cisco in the next year. The bulk of Cisco's $22-billion-a-year business is selling the gear that makes the Internet work--multi-thousand-dollar routers and switches, smaller cousins of the one onstage. The machines make sure that websites load correctly and that e-mail goes where it's supposed to go. "Today," says Chambers, "what we're talking about is the biggest jump in innovation since the router was first introduced 20 years ago." Chambers is so amped that by the time he leaves the stage, he's sweating as if he'd run five miles.

It's been a long while since Chambers has been this excited by anything, and with good reason. The man who rode the Internet bubble to superstardom has spent most of the past three years digging his company out of a hole. At the bubble's peak, in 2000, Cisco's revenues and profits were growing 50% a year, and its market capitalization was the biggest in the world. When the bubble burst, Cisco went from being one of the most profitable and admired corporations on the planet to just another high-tech fallen star with bloated inventories, too many employees, and angry investors. The stock fell so fast and so hard--from $80 to $14 in a year--that Chambers had bodyguards constantly around him. No one accused him of being a crook, like the cheats at Enron, WorldCom, and HealthSouth. But a charlatan? Oh, yes.

Since then, as many investors already know, Chambers has led one of the most impressive turnarounds in Silicon Valley history. Many didn't think he had the spine for the tough decisions he would have to make--builders like Chambers typically have difficulty tearing down what they have built. But since 2001 he has laid off nearly 10,000 employees, written down some $2 billion in inventory, cut 20% of Cisco's 50-plus product lines, redesigned the remaining products so they require fewer parts, and slashed Cisco's supplier base by 60%. As a result, even though quarterly revenues are 17% below their bubble peak, net income is at an all-time high. It will be years, perhaps decades, before the stock sees $80 again; it recently traded at $22 a share. Even so, it has been the best-performing major high-tech stock for the past two years, better than Intel, Dell, Microsoft, HP, or IBM.

More than anyone, Chambers knows you don't build a high-tech colossus by penny-pinching. You do it by taking on hard problems and turning them into hot products that generate double-digit growth. So now that he has repaired his reputation and set Cisco right again, he faces an even tougher challenge. You see, there isn't a lot of double-digit growth left in Cisco's main business. The company already controls roughly 75% of corporate networks, and corporations with Cisco gear aren't running out to buy more the way they did in the 1990s. Chambers doesn't argue with Wall Street estimates that Cisco's core business will grow no faster than GDP for the foreseeable future--in other words, in the low single digits.

This is why he is so excited about the CRS-1. It is not just another box to sell; its existence sends a message to investors, employees, and customers about Cisco's chances of becoming a growth machine again. Chambers may have already conquered the corporate data network, but that's just scratching the surface, he believes. He wants a piece of the vast market for telecom equipment and services, worth $750 billion a year in the U.S. alone. He wants to expand into corporate voice-and video-conferencing networks, and, for your home, he wants to sell you broadband modems along with the gear that hooks your computers together. Would you like to tie your stereo and telephone into your home network? He'll sell you equipment to do that too. By next year he'll even have a line of home telephones for sale.

At the heart of all this activity will be thousands of CRS-1s, Chambers believes. They'll be housed in those windowless central offices that telcos and cable companies operate around the world. Each $1 million--plus router, Chambers says, can handle 100 times the traffic at four times the speed of anything else out there; its software--which Cisco wrote from scratch--is supposed to ensure the degree of reliability that phone companies have come to expect of their old but unshakable switching systems. "What we're attempting to do as a company is actually fairly complex," Chambers says without a shred of modesty. "We're talking about being the product leader in not one or two areas...we're talking about going to ten, then to 15, then to 20." Ma Bell may be dead, but in Chambers's unhumble opinion, Ma Cisco is just getting going.

Chambers's plans sound positively Microsoftian--and they are. He believes that by controlling the routers and switches that go at all ends of every network he can create a "seamless experience" for users. That means whether you are using a cellphone, a laptop, or an office telephone, you can go from your house to your car, to the beach, to the office, to the airport, and to a hotel--across many different networks--and have each network recognize you automatically and grant you the appropriate access. "People fixate on voice over the Internet, for example, because it's free," says Mike Volpi, Cisco's routing chief. "That's interesting. But even more interesting are things like having video telephony whenever you want it; having a single mailbox for e-mail and voicemail at work and at home; or having a single telephone number that follows you from your office to your cellphone to your home, automatically." He says that within a year, for example, you will be able to buy mobile phones that will automatically connect to your wireless computer network at home if the cell connection is bad.

The financial commitment required to build the product that would enable all this--the CRS-1--is unparalleled in Cisco's 20-year history. Five hundred people worked on it for four years, at a cost of $500 million. That's 4% of Cisco's R&D budget dedicated to one product, when for three out of those four years, Chambers was cutting the rest of Cisco's operations. Being in a roomful of CRS-1s makes you feel as though you're on the tarmac at a busy airport. They suck so much power that Cisco had to upgrade the electrical grid in one of its office buildings four times during development and testing. Each machine, fully configured, costs more than $2 million, but Cisco claims it can do things that nothing else can, like transfer the data equivalent of the entire Library of Congress--128 million items--in under five seconds, or handle three billion simultaneous phone calls. Inside Cisco it isn't referred to as the CRS-1 but as the HFR, for "huge fucking router." "People thought we were a little nuts to be developing this product four years ago," says Tony Bates, who shepherded the project. "They said, 'You're biting off more than you can chew,' and they asked, 'Who is going to need all that capacity?' "

It's becoming clear that a lot of people are going to need that capacity. As the price of bandwidth continues to fall, as more people get broadband connections, and as new applications pop up daily, Internet usage has started to accelerate again. After slowing to a growth rate of merely 60% to 70% a year after the bubble burst, bandwidth demand is now expected to grow at close to 120% a year for the next three years, according to a study by Mark Bieberich at the Yankee Group. The Internet bubble made us all learn to doubt such forecasts, yet this time there is a great deal of anecdotal evidence to back it up. Think, for example, of the increased load being placed on the Internet just by digital photos. A standard e-mail consists of ten kilobytes of data. Zap a picture across the network, and you're talking about a file 100 times that size. Add music downloads and online gaming to the mix, and you get reports like the one from Japan's telecom giant NTT, which says that bandwidth usage in Japan is growing 500% a year. If phone companies find a way to offer cable TV programming over their lines, as many are thinking of doing, it is fair to expect, Bieberich says, that "our estimates will turn out to be conservative."

All of that seems to play into Chambers's hands quite nicely. For longer than just about anyone, he has been talking about how the world's voice and data networks will converge into one big system. Now it looks as though that's finally happening. Perhaps the hottest new use of the Internet is to carry phone calls. When Chambers first started gabbing about that nearly a decade ago, the connections weren't reliable enough to justify the switch. Now, for many people, they are. (For more on net phones, see "The Future Is on the Line.")

Chambers and Cisco benefit from convergence in a simple and direct way: The more people use data networks, the more gear they need to manage all the things they do with the network. Historically that gear has been switches and routers sold to corporations to manage e-mail and Internet access. Now that corporations want to use the network to make phone calls, and now that consumers are setting up home networks to do the same, Chambers is expanding his definition of what constitutes networking gear. This is why Cisco is selling telephones. It's why it's selling home-networking gear and broadband modems through its recently acquired Linksys division. And it's why the company is pushing big routers like the CRS-1 on phone and cable companies so hard.

Success should be a slam dunk, right? Wrong. Take a closer look at the new businesses Chambers has launched into; they bear little resemblance to the one he is in now. In selling to corporate customers, Cisco hasn't always had the most sophisticated box; it has generally won by making itself indispensable to CIOs, with a shrewd blend of marketing, technology, and service. Just as the adage "No one ever got fired for using IBM" was common in the mainframe days of the 1970s, "No one ever got fired for using Cisco" is common among CIOs around the world today. That reputation has enabled Cisco to charge premium prices. Its typical gross profit margins--the difference between equipment sales prices and cost--is 60% to 70%. So if the parts in a Cisco router cost $10,000, Cisco sells it for $16,000 to $17,000.

Few, if any, of the skills that have served Chambers and Cisco so well are applicable to running corporate phone systems, selling networking and telephone equipment to consumers, or selling its new multimillion-dollar routers to telephone companies. As Dell has proven, selling technology to consumers profitably has little to do with technology or service and everything to do with more efficiently assembling and distributing the same thing competitors have on the shelves. Selling telephone gear to companies like Merrill Lynch requires a level of reliability that exceeds anything Cisco has attained before. Merrill's corporate data network can go down occasionally, but if traders can't make telephone calls, they lose business. Cisco learned that the hard way last year when it lost the contract to provide Merrill's internal phone system after a Cisco installation went awry. Selling gear to telephone companies requires proving not just that it is as reliable as the existing phone system but that it is on the cutting edge as well.

What's more, every one of those businesses is packed with competitors that won't roll over and play dead for Cisco. "Many aren't even scared of them," says Sanford Bernstein analyst Paul Sagawa. In the home-networking business, Motorola, Siemens, and Thomson supply the bulk of broadband modems; in the corporate phone business, Nortel and Avaya have been supplying the lion's share of equipment for generations and have net phone offerings that are often cheaper than Cisco's. When Avaya picked up the Merrill Lynch contract last year, it issued a press release gleefully touting its win.

Nothing better illustrates how tough it's going to be for Cisco to dominate its new world than its rivalry with Juniper Networks. Juniper, with less than $1 billion in revenues, is a small fraction of Cisco's size, but it has a reputation for producing big, fast, reliable routers. Indeed, since its release two years ago of a machine called the T640, even Cisco execs admit privately that Juniper has become a thorn in their side. Some analysts say that pound for pound, the T640 is as fast as the Cisco's new CRS-1. That's debatable, but the very idea rankles Cisco executives no end. What isn't debatable is this: Thanks to delays in getting the CRS-1 out the door, the T640 has been the fastest, most capable big router on the market for a lot longer than Chambers or anyone else at Cisco would like. In the past six months it helped Juniper steal market share from Cisco and land some high-profile contracts.

The best known is the so-called Gig-Be contract, a multibillion-dollar upgrade of the communications links among the Defense Department's 100 largest command centers. Late last year Juniper won the contract to supply the system's big central routers, worth roughly $100 million. Cisco executives say that the T640 was one of the key motivators for them as they worked on the CRS-1. Meanwhile, what you hear from Juniper is this: "Cisco has always been a company driven more by marketing and sales than technology," says Christine Heckart, Juniper's chief marketing officer. "They tend to hype things before the technology is really available, and that can leave their customers disappointed."

It's not just Juniper that has made Cisco defensive about its core router business, it's customers too--specifically the phone companies. For nearly a decade Cisco has been trying to persuade them to start moving voice calls off their ancient switching gear onto a Cisco-supplied data network. Cisco's rationale then as now was, They already use Cisco gear to provide customers with Net access, why not phone calls too? But Cisco's entreaties haven't just been rejected; they've been met with derision.

One problem was Chambers's tone-deaf sales pitch. While he was trying to sell the telcos his products, he was also making speeches about how their businesses were going to become irrelevant--how their main revenue source, telephone calls, would soon be freely available over the Internet. The telcos didn't like hearing this, and it didn't make them particularly receptive to his sales spiel.

Another problem was that Cisco's gear wasn't nearly reliable enough to handle voice traffic. In fact, some phone companies told Cisco that its routers were barely reliable enough to handle data, much less voice. The idea that they would rely on Cisco equipment in their core telephone business was laughable. "I had a carrier executive tell me two years ago that he knew more about his Cisco router than Cisco did," says Sam Wilson, an analyst at JMP Securities.

The carriers beat up Chambers so much that he has publicly apologized for his arrogance numerous times since the dot-com bust. Serge Tchuruck, CEO of Alcatel, the big French telecom supplier, says the kind of reliability telcos want in their equipment is very difficult to deliver. "There are plenty of examples of companies that come from the data field that say, 'Pfft, this is easy,' " he says. "I do not want to name names, but I think you know who I am talking about."

What does all that competition mean for Cisco's ambitions? Analysts Sagawa and Wilson think the answer is obvious: If Cisco is going to make inroads in all three of its new markets--selling networking and voice equipment to the home, selling voice equipment to corporations, and selling big routers to phone companies--it will have to compete on price, and if it has to compete on price to win market share, its operating and gross margins will fall.

Chamber doesn't believe it. He maintains that the more markets he is in, the less likely that he will have to compete on price. By his reasoning, Cisco has a better chance of selling a company a corporate phone system if it already sells the company routers and switches. It has a better chance of selling CRS-1s to telcos if it is already carrying a lot of their corporate phone and data traffic. And it has a better chance of selling cheap broadband modems if the phone companies make Cisco their preferred provider for home broadband and networking gear.

Says Charles Giancarlo, the Cisco executive who now runs Linksys, which Cisco acquired last year for $500 million: "Nothing we have done in the last five years has gotten as much attention from the [phone companies] as the Linksys deal. How did our acquisition of a small, low-margin business have such an effect? It was because all of a sudden, by selling consumers Linksys products, we were able to talk directly to the phone companies' customers about services they might be willing to pay money for."

If Chambers delivers big growth in all three markets, the payoff will be huge. Annual demand for big, expensive routers like the CRS-1, for example, is growing at 25% and could be as large as $20 billion in a decade, vs. $2.5 billion today. The market for corporate phones, switches, and the like is already $13 billion a year. And the home-networking and broadband market, just in the U.S., could be $5 billion to $10 billion a year. By capturing half those markets--a smaller share than it holds in the router and switch business, the analysts point out--Cisco could achieve a low-double-digit compound growth rate. That could enable it to nearly double in size by the end of the decade.

But it's going to be a long, tough slog, and that means Chambers might want to consider toning down his rhetoric. Back in the Internet bubble, customers, investors, employees--everyone-- couldn't get enough of his grand visions. They hung on every word as if he were high tech's Elmer Gantry. And now, Chambers announces happily, he's back: "In case you can't tell from my voice, I'm having fun again." He's achieved a lot in three years and deserves to have a little fun. But today in a world where executives who underpromise and overdeliver are more in fashion, it may be a while before the rest of us are ready to join the party.