The lessons of Netscape loom large in Marc Andreessen's mind. The company he founded with Jim Clark in the spring of 1994, the company that made him the Internet generation's poster boy and its first fresh-faced IPO multimillionaire, is a phantom limb, lost in the trenches of the browser wars. Yet more than six years later, at the weekly gathering of the executive committee of Loudcloud, the vehicle for Andreessen's second coming, he is still thinking about a $7 million check.

These days, in the industry that Netscape jumpstarted, $7 million is nothing - milk money - but in 1994 the amount was incomprehensibly large, particularly for the small-town kid who had recently pulled up stakes in the Midwest to settle in Silicon Valley. Running around the office, Andreessen showed off the check to each of Netscape's three dozen employees. The responses ranged from awe - "Holy shit!" - to disbelief - "Is that thing real?" Accepting the check was a mistake, Andreessen is telling the Loudcloud executives. (For the many Netscapees who have gathered here, a key motto is "Make new mistakes.") "Jim Clark made a deal with MCI, which committed us to creating software to run its new ecommerce site, an early vision of an Amazon-like shopping mall called MCI Marketplace," Andreessen tells them. "Seven million dollars seemed too incredible to pass up. But what we didn't realize is that we were selling our soul." The check arrived, and the team was jubilant - but then Netscape had to fulfill its obligation to create software to MCI's specifications instead of software that would have been useful to a wide range of customers. "We became their indentured servants," Andreessen says. "A big early customer owns you."

Loudcloud's executive committee meetings are held each Monday. This one is taking place in an unadorned conference room called Notorious B.I.G. (another goes by the name Queen Latifah) at the company's temporary Sunnyvale, California, headquarters. For much of the meeting, Andreessen, who is Loudcloud's cofounder and its chair, seems distracted, checking email on his pager, twirling dials and pumping buttons as if it were a Game Boy. But he doesn't miss a thing. At one point, there is a discussion about whether a manager should be promoted despite his breach of the company's legal policy. Like every technology company, Loudcloud prohibits new employees from bringing any proprietary information from their previous employers, but the manager in question ignored the edict. As a result, a number of executives at the meeting are disinclined to approve the promotion. Says one, "He hasn't even shown any remorse!"

Without turning his attention away from the tiny screen, Andreessen gives his ruling: "Promote him, but give him 30 days in the stockade."

Yet when the subject of Nike is raised, Andreessen tucks the pager into his breast pocket and leans forward on his elbows. The Loudcloud sales team has been negotiating with the footwear giant, which could become the company's first Fortune 500 customer. Jason Rosenthal, head of business development, points out that signing a company of Nike's stature would be an enormous coup, particularly this early in Loudcloud's history, and with its hopes of an IPO early next year.

The members of Loudcloud's research and operations teams aren't so sure. The head of the research department, In Sik Rhee, says that he would first like to know if Nike's technological needs fall within the model the team has designed. If Nike uses technologies that Loudcloud doesn't support, Rhee argues, it would be a mistake to sign them up.

"But isn't it worth stretching for a company like Nike?" asks B. J. Bushur, Loudcloud's sales director. "I mean, Nike!"

That's when Andreessen tells the story of the $7 million check - one of countless moments in which he and Loudcloud's executives revisit the catalog of incidents and insights they carried away from their wild ride at Netscape.

"It all comes down to one thing: whether we are building a burger or a company," he says. "If we are building a burger, we take every customer that comes along and cash the checks as quickly as we can. But if we're building a company, we say no to any company that distracts us from the long-term mission."

There is some head-scratching. "A burger?"

"A burger is meant to be flipped," he explains, "whereas a company is meant to be here for the long term. It's the central issue in this company and it affects every decision we make. It affects who we hire, how we spend our money, how we incentivize our people, and whether we accept a customer that doesn't necessarily fit into our business model." And just in case anyone still doesn't get the point, he firmly concludes, "We aren't in the burger business."

If not burgers, then what? The idea behind Loudcloud is simple, perhaps brilliantly so: to automate the process of building and maintaining Internet sites in order to provide Web-hosting services on an unprecedented (and unprecedentedly lucrative) scale. But the simplest idea is often the hardest to see, and Andreessen's act 2 has been a long time in the planning. For more than two years, even as he held the post of America Online's CTO following Netscape's acquisition by AOL and Sun, he scoured the digital landscape, searching for the perfect business.

The man he wanted at his side, more than any other, was his close friend Ben Horowitz, a popular Netscape and AOL manager. The 34-year-old Horowitz is thoughtful, methodical, and even-tempered (especially compared with the mercurial and occasionally morose Andreessen). He'd run several Netscape product divisions, headed up the company's directory and security product lines, and then managed AOL's ecommerce division - what basically became Shop@AOL, one of the largest shopping destinations on the Internet. "Horowitz was the manager everyone at Netscape wanted to work for," says Scott Dunlap, Loudcloud's head of product management. "The great VPs of Netscape all came out of Ben Horowitz. Ben helped people build careers and pulled together teams better than anyone else."

Even as the other Netscapees scattered to the winds, Andreessen's inner circle - which included Horowitz, Howes, and several others - stuck together; they had bonded in the course of the extraordinary, short life of one company, and now they would seek redemption with another.

Andreessen says he never considered anyone else for Loudcloud CEO, not even himself. "Whatever value I can contribute to the company, they're getting it now," he says. "I would hate the job and I wouldn't be particularly good at it. I can contribute more if I don't have to be consistent. I am free to drive everyone crazy with new ideas and impatience. I'm happy to push people right to the edge. I don't mind pissing people off."

In Sik Rhee made the eureka observation that started it all.

Horowitz, by contrast, tries not to, which is one of the reasons Andreessen wanted to work with him. They powwowed early on about the kind of company culture they envisioned, about trust and respect, and how to motivate employees in a business where talent is scarce and hard to keep. Modeling his management style after that of the man he claims as his hero, Andy Grove, Horowitz sets general goals for the company and encourages managers to set broad goals for their staffs. But whereas Grove is known for an almost ruthless analytic zeal (at Intel, one of his best-known maxims is "If you can't measure it, you can't manage it"), Horowitz is considered empathetic as well as effective, which may be why he is described by one employee as "the kind of person you would follow into a pit of fire."

As late as September 1999, when Andreessen resigned from AOL, he and Horowitz still didn't know exactly what they were going to do. But they knew where to find the talent to do it. First Horowitz tapped Rhee, the wizardlike engineer who'd been CTO of the ecommerce platform division at AOL, to serve as head of Loudcloud research. The Korean-born Rhee had also come to AOL via Netscape, after it acquired Kiva, another legendary startup. Rhee cofounded Kiva and designed its main product, an application server, the kernel of which still runs Web sites for E*Trade, Bank of America, Lucent, and others.

Then, in October, Tim Howes came on board as CTO and VP of engineering. A respected specialist in large-scale, high-performance Internet infrastructure, Howes authored LDAP, the Lightweight Directory Access Protocol, which is the Internet standard for directories. Loudcloud's founding team was complete. "It's a team of Internet stars - people who built some of the most impressive technologies on the Net," says Joel Yaffe, an analyst for Giga Information Group.

Throughout the winter, the four met around a table under the circle of redwood trees in the front yard of Andreessen's Palo Alto home. Dogs and cats underfoot, they batted around dozens of ideas for Internet startups. Each idea had a drawback. And because of the complex infrastructure requirements of state-of-the-art Web sites, it would take months to get any one of them up and running. "Before Netscape, 'months' was fine," says Horowitz. "Now, with the frenzy of development of new Internet companies, months could kill you." The group decided to do the groundwork first, so that they could simply drop in whatever idea they ultimately settled on.

It was Rhee who made the eureka observation, drawn from the experience of building Shop@AOL. Rhee and the others on Horowitz's team had been charged with bringing most of the popular Internet shopping sites onto AOL, offering them access to the enormous AOL user base. Eventually, that meant more than 300 merchants, from small online stores to Amazon and eBay. When a company went live on AOL, its traffic increased by a factor of 10 to 100, and the site often collapsed as a result. Crashing sites were unacceptable, not least because AOL's users naturally tended to blame AOL for these problems. Rhee had started researching potential merchants to determine the scalability of their sites. "Almost all of them were rickety systems," he recalls, "doomed from the start." He worked to jury-rig their sites so that they could scale up when they went on AOL.

Now Rhee suggested that Loudcloud toss aside all its ideas for Net companies and instead become the company that put other people on the Net. "I became excited about the idea," Andreessen says, "when I saw that instead of running one of the stores on the block, we'd own the block - we'd run the block." There were many good reasons for companies to pay tens and conceivably hundreds of thousands of dollars a month for someone else to do the technical dirty work. Loudcloud could get a Web site up faster - in a third of the time, Rhee and the others calculated. It could do so much more cheaply, since a company generally needs a dozen or more highly skilled technicians to do the job. But the most compelling argument for a one-stop shop for Web infrastructure was that companies lose focus on their business if they have to continually worry about their sites.

Web infrastructure, or Web hosting, was hardly a novel idea: There are hundreds of companies - from LogicTier and MimEcom to Intek and USinternetworking to AT&T and Siebel - that help other companies go, and stay, online. Like the rest, Loudcloud would lease space at data centers, setting up hardware, network, and storage systems. The difference - and the potentially enormous competitive advantage - would lie in a new level of automation.

"The dumb way to do this business would be to try to do what other companies do," Andreessen says. "For every customer, we could hire another 10 people to get them up and run them. It would work. But at the end of the day, there wouldn't be that much value. The business would only scale by the number of bodies you could hire, which means that your quality of service would degrade over time. We wanted to do this for hundreds and then thousands of customers without having to hire that many new bodies."

Which is where Tim Howes came in. He spearheaded the creation of Opsware - software that makes it possible for Loudcloud to put up a new Web site with a keystroke or two. Dedicated servers are automatically configured with standardized operating systems, storage devices, load balancers, firewalls, security systems, database software, and more. Once the site is running, Opsware monitors it on a 24/7 basis, "grabbing" new servers and replicating the configuration and databases - scaling up to deal with increased traffic, and adding whatever new capacity is needed, whenever it's needed, over the network.

Customers start by choosing from a menu of Clouds: a Database Cloud, an Application Server Cloud, a Mail Cloud, a full Web Cloud, et cetera. They pay a monthly fee, based on the capacity they need, but Andreessen notes that Loudcloud doesn't get a cent until a site is up and running, and that clients are guaranteed 100 percent uptime. If a client wants to add services - say, a shopping basket or a secure payment system - Opsware can drop in any number of new Clouds that can be switched on and off. If a new software version comes in or a new feature is added to Loudcloud's menu, it can be installed automatically on the systems of the customers that want it.

Tim Howes spearheaded the creation of Opsware to speed-build sites.

In a meeting with Loudcloud adviser Mike Homer - a Netscapee, investor, and close friend of Andreessen's - a pair of engineers shows off the new piece of hardware that allows the Opsware software to monitor every piece of equipment in the Loudcloud system. Through a barcode connector, each server and component will identify itself to the other components and to the Opsware that monitors them all. One of the engineers presses a button on the scanner and a wireless signal is sent to a computer that instantly identifies the component. The scanner is a pretty cool-looking toy, and Andreessen hogs it. When someone asks what he's doing, he replies, "Asteroids."

Loudcloud isn't trying to reinvent the wheel; it's using its expertise to select off-the-rack programs (from Sun, Microsoft, IBM, and Oracle, among others), design new ones when necessary, operate them in various combinations, and run them in a high-volume Internet environment. Scott Dunlap predicts that the venture will make existing Web-hosting operations "seem like the pre-assembly-line car industry."

But the model has its perils. "The [Opsware] software distinguishes them," says a manager of IBM's Web-hosting division in Europe. "But one wonders if it's too soon to push a standard. It could be the basis of their success or their failure." Rick Juarez, an analyst at Robertson Stephens, who is continually trying to explain the enormous sector of Web-hosting and other Internet infrastructure companies to the financial community, thinks the approach has promise: "If Loudcloud successfully pulls off this highly engineered approach, it will be a breakthrough. The chief risk is that they will try to be too sophisticated, and it will become too complex to use and too onerous to manage."

Benchmark Capital invested $15 million in Loudcloud, after the founders put up $8 million (the bulk of it from Andreessen) and before they secured $45 million in debt financing from Morgan Stanley Dean Witter. When Benchmark lead VC Andy Rachleff ticks off the reasons his firm anted up, he mentions the quality of the team, the enormous potential market, and the fact that Loudcloud's revenues should ramp up much faster than those of its competitors. "They're unique in that they're the first to allow a company to completely outsource the Web-hosting process," he adds. "Some manage up through the operating system, but Loudcloud manages through the application layer." He praises Horowitz for creating a productive management culture, and for being selective about Loudcloud's customers: "They want to be sure the customer is going to grow and provide more revenue."

Given the recent spate of explosive viruses and denial-of-service attacks, Loudcloud also wants to be sure its customers are protected. "Security concerns cut across everything we do," Howes says. "From the beginning, we had to consider it in every level of our offerings, from the servers to the application layers. The more sophisticated the black-hat community becomes, the more sophisticated we become." He says the all-important Security Cloud, part and parcel of every other Loudcloud service, uses off-the-shelf security technologies - such as Cisco's firewall system and NetScreen for the VPN connections - along with some proprietary additions, and includes complex password and intrusion detection systems. It got its first test when a hacker tried to shut down a Loudcloud customer's site as part of the broader attacks on Yahoo! and eBay et al. "The breach was detected and denial of service failed," says Howes.

When Andreessen first described Loudcloud's business to the press in January - "It's like providing the electric power grid, and companies can plug in" - it sounded pretty unsexy. "It is unsexy," says Giga's Yaffe. "It's meat and potatoes." Opsware isn't inherently dramatic; it's a way to speed and scale up what is now a difficult, labor-intensive process. Indeed, if Loudcloud's founder had been anyone but Andreessen, the company would not be getting anything like the attention it has received. But meat and potatoes? The world expects flash and innovation from the man who defined Netscape Time, who has been compared - persistently if erroneously - to Steve Jobs and Bill Gates, who appeared on the cover of Time barefoot and wearing a crown. Those expectations are Andreessen's blessing, and so, of course, they are also his curse.

Andreessen dislikes - no, dislikes is too weak a word - he loathes being asked about his background, and has walked out of interviews when questions about his family or his hometown (New Lisbon, Wisconsin; population 1,491) have been raised. When pressed, he becomes agitated, then sullen. Finally, he offers a feeble explanation. "I'm an introvert," he says. "I would hide if I could get away with it."

As disingenuous as that sounds coming from a highly visible, sometimes cocky industry leader who once called himself a "media whore," his reaction to personal questions hints at a real discomfort. Over the years, a relatively standard-issue geek profile has been constructed in the thousands of pages devoted to Andreessen. His mother, Pat, was a customer service representative at Lands' End, the catalog company. His father, Lowell, was a sales manager for Pioneer Hi-Bred International, a seed company. At his tiny high school, with 240 students, sports were the measure of a boy's worth, and Andreessen was both unathletic and uninterested. He learned to program from a book he read when he was in the hospital recuperating from an operation. On a computer in the school library, he built a calculator that could do his math homework. That night, when the janitor turned off the power to the building, the program disappeared.

His teachers and classmates remember Andreessen as a good student who liked computer science and math, as well as English and history. Ken Adams, who was principal of New Lisbon High when Andreessen attended, told Fortune in 1996 that "Marc had an intellectual capacity that could intimidate people." And classmate Clay Haschke said, "Marc just seemed to have a broader sense than the rest of us. Say we were talking about God or something. He would talk about it in a more complete way, with more than one view. And we'd sit back on our heels and say, 'Wow, oh yeah.'"

Reporters searching for Andreessen's backstory have been stymied by his silence on the subject, and more than one article has noted that since he left home, he has had little, if any, contact with his parents. But he has racked up an impressive list of mentors - father figures? - starting with Jim Clark.

"Yahoo! didn't have anything to lose," Barksdale notes. "We did."

The path that led Andreessen to Clark is one of the Internet's most familiar tales: In the early 1990s, a towering, round-faced, 21-year-old engineer with short hair and short pants was studying computer science at the University of Illinois at Urbana-Champaign and writing Unix code at $6.85 an hour for the school's National Center for Supercomputing Applications. He decided to devise a simple interface for all the functions of Tim Berners-Lee's World Wide Web. With Eric Bina, a fellow NCSA programmer, he spent three months of nights and weekends creating the original version of what they called Mosaic.

Andreessen now says that his goal was to democratize the Web - "Why shouldn't everyone have access?" - though he doesn't pretend to have anticipated the world-altering results. "The work was fun, but no one was taking it seriously," he says. "Everyone was looking at interactive television, which was going to be pushed at first by a single killer app: movies on demand. The future of the world was going to be determined by AT&T and TCI and Time Warner. Every household was going to be hooked up." The Internet? "It was thought of as a ploy - a low-bandwidth ploy," he says. "It was nerds and scientists and typing. All that crap. That's what everyone thought - Microsoft and everybody else. I just thought, 'I may as well work on this now, and then when I get out of college I can go work for Silicon Graphics or Time Warner or TCI.'"

Instead, he went to work at a small software company in California, where he received an unsolicited email from Clark, the already famous founder of Silicon Graphics. Clark, who had heard about Andreessen and Mosaic, was fishing for new business ideas. The two met and discussed, among other products, a Nintendo online gaming service. One night, after several bottles of burgundy, Andreessen uttered the words that have since been reprinted in every telling of the Netscape saga: "We could always create a Mosaic killer - build a better product and build a business around it.''

Clark raided Urbana-Champaign for Bina and other programmers handpicked by Andreessen, and then moved them to Mountain View, where he set them up in an office suite in the City Center building. Following the arrival of new CEO Jim Barksdale, formerly of McCaw Cellular and Federal Express, in dazzlingly short order Netscape became the fastest-growing software firm in history, and its browser, Navigator, emerged as the key to the Internet boom. Netscape's unprecedented success caught the technology industry off guard, but the major players - including Sun, Oracle, SGI, and, of course, Microsoft - soon saw the writing on the wall and switched their focus from interactive television to the Net. In addition to galvanizing existing technology companies, Netscape launched literally thousands of Internet startups. The Netscape IPO, in addition to making the team rich (Andreessen was suddenly a multimillionaire and Clark was well on his way to becoming the first Net billionaire), unleashed the headlong rush into the public markets. Ask the founders of any Internet company what inspired them, and there's a good chance they'll answer, "Netscape."

Going public as Netscape did, before making a dime in profits, was a revolutionary act. So was Netscape's business model. Andreessen says that both Clark and Barksdale were easily convinced not to charge individual users for Netscape's famous browser. How else would Navigator trounce Mosaic? Revenues would come from Web site building, ecommerce, email, and server software, as well as browsers for corporate customers and service contracts. The company grew quickly - after a year there were 1,200 employees - and brought in startlingly high returns. Andreessen recalls, "In 1995, from Q1 to Q4, Netscape's revenue went from $5 million to $10 million to $20 million to $40 million. No one had ever seen anything like that. It was faster than Lotus' first year, which was the previous record in software. Shit! That's it! We are a hit!"

Revenues reached $100 million in 1996, which is when Microsoft began waging the browser wars in earnest, and other major software companies began to attack Netscape as well. A Fortune profile of Andreessen raised some prescient questions. "Is this 25-year-old programmer who never took a business course in his life an asset to his company? Can he help Netscape continue its phenomenal growth at a time when its market has been invaded by Microsoft? Is he anything more than a one-hit wonder, the Golden Geek of the moment?"

By the start of 1998, Netscape had hit a wall. It reported an $88 million loss, while its share of the browser market had dwindled from 80 to 60 percent. (Today it's less than 40 percent.) Analysts who had heralded Netscape as the most important new company in history were now sounding its death knell. As CTO, Andreessen had been somewhat insulated from day-to-day management, but six months earlier Barksdale had promoted him to executive VP in charge of Netscape's 1,000-person product-development group. Andreessen couldn't stop the implosion. Maybe no one could have. The company's role in instigating the Justice Department's investigation of the browser wars, and the resulting antitrust suit against Microsoft, turned out to be Netscape's greatest legacy - and the greatest factor in creating the widespread public perception that the company's demise came at the hands of Bill Gates & Co.

But these events obscured a critical fact: There was one road not taken by Netscape. While Andreessen and his colleagues were battling Microsoft and other technology giants for a piece of the enterprise software market, they failed to exploit what turned out to be the most important asset in the Internet world they had done so much to create: Netscape's Web site - for a time the most popular site on the Net - where people went to download the browser.

By now it's a favorite call of Monday morning quarterbacks to point out that Netscape didn't take advantage of its Web site's huge traffic even as fledgling Internet companies were fighting viciously for all those eyeballs. "Yeah," Andreessen admits, "we could have been Yahoo! - if Netscape was to become a $100 billion company, that would have been how to do it. There was a point in time when that could have been done." Instead, Netscape gave Yahoo! - as well as Excite, Lycos, and others - access to its traffic for a relatively small fee. By the time Netscape entered the media portal business with a refocused Netcenter, in the summer of 1998, it was too late. The company that launched a thousand startups and woke the world to the power of the Web had ended up looking the wrong way.

Scott Dunlap joined by quitting E.piphany on the day of its IPO.

Andreessen recalls the genesis of this fatal miscalculation. "In a war, you have a battle plan up front," he says. "You get into the war, and shit is happening all over the place - people are being blown up. You have to make decisions about what to do with incomplete information. Looking back, though, given the same information at every point in time, I would have made the exact same decisions over again." At the time, he notes, "Everybody in the Valley thought that Netscape had a clearly superior business to Yahoo! - financials ultimately drive almost everything a company does, even companies that are idealistic and try to pretend that is not the case. Our revenues had been doubling each quarter. I don't know how many people would have switched the model at that point."

True enough. The only hitch in this logic is that Marc Andreessen did see how the Net would change the world. In speeches, he described the digital future with uncanny foresight. But he didn't push his vision on his older, more experienced colleagues. Quite the contrary. And this fact speaks volumes about who Andreessen is - a very different person from the image created by his Netscape stardom.

For one thing, it explains why the comparisons to Gates and Jobs are inaccurate. Andreessen has always been a man of many mentors. Since he was thrust into the limelight by Clark, Andreessen has had the backing of some of the industry's heaviest hitters, including John Doerr (of Kleiner Perkins Caufield & Byers, the VC firm that backed Netscape), Jim Barksdale, and Peter Currie, then Netscape's CFO. "Barksdale influenced Andreessen into wanting to become a businessman," says one longtime observer. "A good manager, to be sure, but a businessman rather than what Andreessen is innately: a revolutionary thinker. Marc wasn't able to be both." (Barksdale also radically improved the punctuality of his young protégé, by locking him out of one of Netscape's morning meetings.)

While the world was lionizing him as a Gen-X icon, Andreessen was telling Fortune that "the stuff that interests me is business strategy and customers.... There's a triangulation between what's possible with the technology, what business needs people have, and what's actually practical. You triangulate on that, and that's where the sweet spot is for making a successful business and driving it further." That's a sharp contrast to someone like Steve Jobs. When Jobs brought in Pepsi CEO John Sculley to run Apple, the almost instantaneous conflict between Jobs' visionary drive and Sculley's practical-minded company-building led to Jobs' exit.

Instead of pushing for Netscape to become a Yahoo!-like media company and portal, Andreessen was following the lead of Barksdale, Currie, and Doerr, trying to create a bigtime software company - and not a sexy software company, either, but one that focused on the intranet infrastructure of corporate computer networks. "There was money in this, there was no money in that," Andreessen repeats. Of course, Jerry Yang and David Filo proved him wrong, but as Jim Barksdale points out, "They didn't have anything to lose. We did."

Still, it remains one of the central ironies of the Netscape saga that Yahoo! would finally disprove the theory that no one could beat Microsoft at a business Bill Gates wanted. Compounding the irony is the fact that Yahoo! built its audience on Netscape's back; and that, in the Web's early days, when Andreessen was a budding celebrity and Yang and Filo were penniless Stanford grad students running a cultish directory service from which no one could see a proper business being built, Netscape passed up the opportunity to buy Yahoo! for a song.

To this day, Andreessen, like Barksdale, defends Netscape's accomplishment as worthy of admiration and respect; for him, the bottom line is that "in a little over five years, Netscape created $10 billion in new wealth out of nowhere. In particular, anybody who ever bought a share of Netscape stock and held it made money. Name another Internet company for which that is true! There is none. And two, Netscape made Loudcloud. That's all I need to know."

Andreessen faced another problem: As the Net's first Rich Kid, he bore the weight of the media's need to create stars, and suffered the backlash that inevitably follows in the quest for a different story line - a backlash that was heightened, perhaps, by the fear and jealousy of aging baby boomers. Almost as soon as Timehad literally crowned Andreessen, the efforts to debunk him started: He hadn't really had the idea for Mosaic; he'd stolen credit from other programmers; he was a fraud. The accolades for his Netscape achievement were thrown into question, then Netscape itself was gone. His first professional experience hadn't been a simple success, but a success-disaster.

Given the drama of Netscape's rise and fall, the comings and goings of its executives were followed with the rapt attention once reserved for the love affairs of movie stars. Jim Clark famously pursued the new new thing (the latest: Shutterfly). And CEO Jim Barksdale left soon after Netscape was divvied up by AOL and Sun, launching a VC firm with former CFO Peter Currie. But the wunderkind inventor of Navigator mystified many by remaining at AOL as CTO, while Netscape dwindled into a glorified R&D wing of a mass-market service and content provider.

"Ben takes every detail seriously," holding weekly one-on-ones.

To those close to Andreessen, his departure was only a matter of time. Andreessen insists, however, that he had wholeheartedly committed to AOL: "Committed? I committed! I bought a house in Virginia! No one else did that!" He was intrigued by AOL, he says, because "more than any other company, AOL is in a position to bring the Internet to the rest of the world. They are the ones who will double, triple, quadruple, and quintuple the number of people on the Net. I wanted to stay on and see what I could do."

There are varying reports about how well Andreessen and Steve Case got along. During negotiations for the AOL-Sun acquisition of Netscape, Andreessen presented Case with a report on the future of AOL that included a list of the "top 10 threats" to the company. (He and Horowitz now keep a changing list of the top 10 threats to Loudcloud: At the moment, "growing too fast" is right up there.) Case was impressed, but according to Rob Enderle, another analyst at Giga Information, "Case didn't know what to do with Marc. He gave him the title and a lousy office, and that was about it." Barksdale sharply disagrees, saying, "Steve saw him as an important asset."

Andreessen is short on the specifics of what he accomplished at AOL - he mentions "some decisions about technology" and the acquisitions of the Internet music startups Spinner and Winamp - and says his main goal was "to figure out ways to help them build on what they had." It may be that he was biding his time, torn between fresh ambitions and the lingering wounds that ambition had already inflicted. Meanwhile, Case was plotting a new course for AOL: the merger with Time Warner. "Knowing Marc, being lost in AOL Time Warner would have been akin to being burned alive," says Mike Homer. Ben Horowitz adds, "Marc has big ideas and wants to see them enacted yesterday. You can't do that in a big company like AOL."

"When I project five or ten years down the road, I want to look back and realize that I was at the creation of something great," Andreessen says now, and it's clear he doesn't consider Netscape to be that legacy.

Present at the creation, Loudcloud's founders have since reeled in some of the Valley's hottest talents. Jason Rosenthal, Horowitz's director of product marketing at AOL, had quit his job and was planning to spend two months mountain climbing when his former boss told him about Loudcloud. After meetings with the other three founders, Rosenthal was summoned to Andreessen's house. He was nervous, and taken aback by Andreessen's first question: "How is your sense of smell?" Rosenthal said he thought it was pretty good, whereupon Andreessen explained that his is very weak. The chair of Loudcloud had a date that evening and was convinced that one of his bulldogs had had an accident in the house. He asked Rosenthal to try to find it. Followed by Andreessen, Rosenthal sniffed about, but he couldn't detect anything malodorous, and the two got down to business. Before the afternoon was over, Rosenthal had agreed to join Loudcloud as director of business development and marketing.

Andreessen took Chuck Katz, of the law firm Perkins Coie, which represented McCaw Cellular and Amazon (he's also the husband of Netscape's former general counsel, Roberta Katz), out to dinner and persuaded him to sign on as Loudcloud's chief counsel. "The original Netscape browser that made the Web accessible and that Microsoft copied in a far less elegant way is a reflection of Marc Andreessen's brain," says Katz. "It is incredibly complex and yet simple, a portal to where we want to go. You don't meet people like that very often. When presented with the opportunity, you want to be in the same stadium." Over a beer, In Sik Rhee pulled in Jonathan Heiliger, who at 23 has quite a résumé - Global Crossing's GlobalCenter CTO; founder and senior VP of Frontier Internet Ventures - and is considered one of the most innovative operations executives in the business.

But the greatest gesture of faith in Loudcloud's future may have come from product management VP Scott Dunlap. Dunlap's two years at E.piphany were about to pay off handsomely; the company was going public in two weeks. Then he met with Andreessen. On the day of the E.piphany IPO, Dunlap gave notice. "I can teach a course now in How to Ruin Your CEO's Perfectly Good Day," he says now. (At the meeting in which he announced he was leaving, one colleague hurled a telephone at him.) The move cost him millions of dollars, but, he says, "I saw the light in Marc's eyes."

When Rosenthal was told to sign up Loudcloud's first customer, he said he could do so within six weeks. "What takes six weeks these days?" Andreessen responded. Rosenthal landed Acteva in a week, and in the next few months a half-dozen small to midsize Internet companies signed up. Chris Wong, the CEO of SkillsVillage, an early customer, says that his company had put up its own Web site, but it was a fiasco. "I was losing sleep," he says. "And it wasn't what I wanted to spend my time on." He knew Andreessen - who is an investor in SkillsVillage - and had heard about Loudcloud. He says he jumped at the opportunity to do business with Andreessen's company: "Nothing like being able to hire the architects of the Internet to make sure your site goes up and keeps running."

But even as the talent and the clients keep rolling in, there's another lesson from Netscape that keeps Andreessen and Horowitz up at night: trying to get around the Law of Crappy People. "It scares the shit out of me," says Andreessen. "The law applies to every company that gets big," says Horowitz, "especially companies that get big fast, so we're a prime candidate.

"All you have to do is hire one person who isn't very good," Horowitz continues. "The Law of Crappy People kicks in because the worst employee at any level becomes the de facto standard for that level. Your executives sit around the table. Some EVP wants to promote one of his or her directors to vice president. Or maybe they want to bring in someone new. Someone at the table says, 'That person isn't good enough,' to which the first person responds, 'Hey, you've got Joe Schmo, who is a bonehead. This guy is better than Joe.' So the guy who hired or promoted Joe Schmo shuts up."

Andreessen says, "You will inevitably make a mistake. The minute you do, the quality degrades. A people hire B people, and B people hire C people. So the bad people breed like rabbits - they hire more people like themselves or worse. We saw it happen at Netscape when we ramped up early on."

"I don't like risk," says the man who inspired a generation to take risks their parents never dreamed of.

To avoid the inevitable, Horowitz and Andreessen are trying some unconventional tactics. One of the first employees - number 9, in fact - was a recruiter, Jeanne Romano. She'd had no interest in an office job, but loved the idea of building a company "from the DNA." "People in-house have more skin in the game," she says. After three months, 3 of Loudcloud's 100 employees were recruiters; now there are 7 on staff. Similarly, though business schools teach their students that an HR department should be in place by the time a company has 150 employees, Loudcloud hired an HR person when there were only a dozen.

Six months after it was founded, Loudcloud had nearly 200 employees. There will be 300 by this fall. In October, the operation moved from Andreessen's home to a grim office on Independence Drive in Menlo Park with a long line of Costco folding tables and workstations plugged into a daisy chain of extension cords. Ray Suorsa, another highly regarded engineer who signed on to work with Howes' team, says, "It was the most dangerous place I've ever seen. Each time a new employee arrived and plugged in a computer, everyone would take a breath to see if the whole table would go down."

When the Independence Drive office was bulging at the seams, Loudcloud relocated to an equally nondescript space on Tasman Drive in Sunnyvale: several rooms filled with quartets of cubicles and the standard tech company complement of Nerf guns. Jonathan Heiliger's NOC, or network operation center, crowded with a dozen computers, was described by one of its occupants as "a modest hell." Andreessen and Horowitz shared an office, with a tiny refrigerator stocked with healthy snack foods on Marc's side, and, on Ben's, a glass-encased Oakland Raiders football helmet signed by the entire team - a gift from the Morgan Stanley Dean Witter bankers who hope to handle the Loudcloud IPO.

"An environment in which you are breathing down each other's necks is incredibly exciting," Andreessen says. "You don't want to lose the energy of all those people creating and complaining. At Netscape, after all the whining about doubling up in cubicles, being packed together, we moved into the fancy, spacious buildings on the Netscape campus. Everybody had big cubes, and it was all spread out and everybody had things exactly the way they wanted. The buzz was gone, the energy was gone. The lesson: Pack them in to the point where they're just about to riot."

But the Tasman office couldn't hold Loudcloud for long. And in late May, despite Andreessen's belief in cramped creativity, the company headed for a roomier, more permanent home.

Andreessen drives a monster Yukon (when he's not driving a monster Mercedes), which has a copy of Dr. Spock's baby book on the passenger seat. Before the book can be misinterpreted, he explains - emphatically - that it was a gift from a potential Loudcloud customer, drspock.com. "I had better get rid of it so that no woman gets the wrong idea," he says, as he heads away from the Tasman office.

Though Andreessen has lived and worked in the Valley for most of a decade, he overshoots the freeway exit for his destination. The Yukon glides off at the next ramp, which happens to be the one that leads to the corporate campus of Netscape. Andreessen turns into the parking lot and the car weaves past a half dozen Netscape buildings. "It's a ghost town," he jokes. "They rent out all these cars and park them in the lot so that it appears that there are people in there. I went inside recently and there's no one - it's empty."

After getting back on the freeway, Andreessen pilots the Yukon to Loudcloud's new headquarters, a scaffolded, three-story, green-and-cream building that he calls "the Taj." Though he seems proud of its sprayed stucco exterior and jade-colored tiles, the building makes him nervous. "You're doing great and then you make the key mistake: You build a monument to yourself," he says. "We did it at Netscape. The Edifice complex worries me because it has marked the beginning of the end for many companies. I worry that employees will see this building and think, 'Man, we've made it.' You've never made it."

This Netscape lesson notwithstanding, Andreessen is resigned to the Taj, partly because more space is needed, partly because "it's not real expensive, given the Valley," and partly because he knows that Loudcloud needs the right public face. The lobby has a comfortable waiting room that shows off the NOC, which is visible through portholes. The NOC guys have traded their modest hell for some "très élégant" new digs, because, as Heiliger says, "customers like to see bits flying across big, flat-panel screens."

These and other details were on the agenda at a late-spring meeting, just before the big move. Dave Conte, who designed the Taj, arrived with rolls of plans and carpet samples and unloaded them on the conference table before presenting the executives with a heady problem: "How fancy of a lobby is too fancy and how fancy is not fancy enough?"

Choosing carpeting, tiles, and other features was particularly tricky, since the Taj costs had risen. "We are talking about spending $5 million to move in, for everything from phones to security to chairs," Conte said. "That is a sizable investment - in Netscape territory," added Horowitz. "It's a lot, but it may not be outrageous for a company that plans to be around awhile. This will be our headquarters as we grow."

Jonathan Heiliger wanted to know if the new conference rooms would have names. "Let's stick with rappers," Horowitz replied.

"How about dictators?" asked Scott Dunlap.

There were suggestions from around the table: Kabila (as in "management training in Kabila"). Khaddafi. Idi Amin. Pinochet.

Then Jason Rosenthal weighed in. "How about serial killers? All-hands meetings will be held in Manson."

"Why don't we name them after Supreme Court justices?" asked Chuck Katz, the general counsel, effectively ending the debate.

At the end of the day, in the dining room of the Lion & Compass, Andreessen and Horowitz ordered only appetizers, since both had dinner meetings that night. Then Loudcloud's chair and CEO took a moment to reflect on their latest triumph: signing a deal with Nike. "We did it on our terms," said Andreessen. "Everything we're doing for Nike is applicable to all of our customers. We waited to take on someone of Nike's size, but we're proving that we're ready to ramp up to other big customers." Within weeks, four more customers would sign on: Echo Online, Gamers.com, Mercury Interactive, and OneClip.com. By June, the list had grown to include Acteva, SkillsVillage, HomeGain, CFOWeb.com, Catapulse, Wish.com, Dunlap's alma mater E.piphany, drspock.com (of the baby book), ZapMe!, StatementOne.com, and Epoch Partners.

Along with the high-fiving, there was an earnest discussion of management techniques. Horowitz expounded on the precise, careful system he has put in place at Loudcloud: He holds weekly one-on-ones with his managers, and whereas most companies give performance reviews once a year (and even those often fall by the wayside), he insists that every employee be reviewed - and review his or her manager - every 90 days. "These days, who can even remember what happened six or nine months ago?" he asked.

Andreessen listened approvingly. "Ben takes every detail seriously because he genuinely wants to build a company better than anyone has ever built one," he said. The day had been spent on the less glamorous aspects of starting a company - such mundanities as lobby tiles and the frequency of performance reviews - but the notoriously impatient Andreessen appeared to be in his element.

Andreessen is speaking at a conference for Indian entrepreneurs, at 2 pm at the Fairmont Hotel in San Jose. Though the hour is approaching, he ignores the voice of his Mercedes' in-dash navigation system and detours to the Taj - he doesn't want to get to the hotel too early. "I hate schmoozing," he says. After the drive-by, he realizes that he will still arrive at the hotel with minutes to spare, so he reverses direction and heads for his home in Palo Alto, where he checks on the two bulldogs and Australian silky terrier in the backyard. (They have their own fountain and a cushy bed built into the concrete fence.) "I have a surprise for you," he teases the dogs. He's flying to Los Angeles after the speech to retrieve a third bulldog.

When he was at Netscape, Andreessen was a famously bad public speaker - "excellent for insomniacs," in his own words. But after some coaching, he can hold an audience with a well-articulated vision and a few well-rehearsed jokes. Driving to the Fairmont, however, he spends much of the time talking about how best to cancel. "The difference between me now and a few years ago is that I would have canceled then," he says. "Or just not shown up."

It's a quick speech and a resounding success. Andreessen quotes Sun chief scientist Bill Joy, describing the evolution of the Net over the next five years as "not one Web but six Webs" - PC, wireless, interactive TV, telephone, embedded (in appliances), and business. The Internet, he tells the crowd, is "the first medium based on software. So it's the first medium that can continue to evolve and adapt and mutate." Pulling out his RIM pager, he makes the audience laugh by predicting that in the future, "humans will have very long and very narrow thumbs with many more joints. Now I have carpal tunnel syndrome, but only in my thumbs." He ends with a plug for Loudcloud, of course: "AOL, Amazon, Yahoo! had to do it for themselves. Now we can provide it for everybody." Afterward, he is swarmed by audience members who pepper him with questions and business cards. A number of them want to sign up with Loudcloud.

The burger-eating, moon-faced kid who shook the tech industry has become, at 29, an elder statesman. He can (and does) talk tech with everyone from Larry Ellison to Al Gore to Nasdaq president Alfred R. Berkeley III to AOL board member Colin Powell, and he's accruing some of the trappings of moguldom. Witness the new house in Malibu, where he parks his latest toy: a Hummer that belonged to Arnold Schwarzenegger, for which Andreessen bid $140,000 at a charity auction last May. When he walks into a room, the energy in the space increases, as if someone had cranked up the volume, treble, and bass all the way. He's fit (thanks to a workout routine and those raw veggie snacks) and polished, with a sharper wardrobe and a less-sharp tongue.

During a disastrous presentation from the representatives of a PR company, for instance, it becomes obvious that Andreessen would like nothing more than to lunge across the table and choke someone. Instead, in a measured tone, he tells them, "Next time you'll do better with us if you bring us choices instead of presenting us with one option." Later, Mike Homer cites this as an example of how his friend has changed. "He could have destroyed them, but he chose not to," Homer says. "In the old days, he would have."

Jim Barksdale, who had to cover for some of Andreessen's brasher comments about Microsoft, lamented during the trial that his protégé "often says things that I wish he wouldn't." But today, Barksdale notes, "it's so impressive to see how much he has grown - as a businessman, as a person." Barksdale can rest assured that Andreessen has - at least in public - perfected the art of circumspection when it comes to his old nemesis. Asked about the Microsoft antitrust trial, Andreessen says, "I am satisfied that the court has been able to handle this. It is really hard for me to have an opinion at this point. Part of it is that I am conflicted - I am a partner of Microsoft now."

Most people who have been with Andreessen from Netscape to AOL to Loudcloud remark on his maturity. Homer calls it "studied restraint," adding that "it's still hard for him when he sees the way things should be but aren't. It is almost as if Marc is standing in the future. It is as if he is compelled to act because of what he sees there. It is a little like a Twilight Zone episode."

Though Andreessen acknowledges that Loudcloud's business is "picks and shovels," you wouldn't know it from the rock-star atmosphere that surrounds him. As Chuck Katz notes, "Whatever Marc does is electrified - it just is. Even if he's not setting out to change a paradigm, it happens. It has to do with the compression of time." Katz may be right. If Andreessen says that time is shifting again, hang on. Netscape Time, or Internet Time, as defined by Andreessen, originally described the time frame in which Netscape delivered new products, which was every six months, phenomenally faster than anyone had done before. Now Internet Time has come to define the speed of life in the age of the Internet - how quickly a company must move if it is to succeed. Scott Dunlap, remembering his first meeting with Andreessen, says, "Marc recounted how the Internet has changed the concept of time. Then he showed, step by step, how Loudcloud is going to again compress time, and not by a trivial amount. Time will change again now, this time by a magnitude of three to five. Well, here was the guy who started the whole thing and he was back for more. Another paradigm shift was coming."

Given Loudcloud's substantial (and mounting) competition, there's no guarantee that Andreessen will end up spearheading whatever paradigm shift his company represents, but there's certainly a large pie to be divided up. Dataquest forecasts that the broadly defined Web-hosting market will be worth $5.9 billion by 2003, up from $1.6 billion last year. And, in its arena, Loudcloud has a lot going for it: a chair who's in peak form, a CEO with unusual management skills, and the software expertise of Tim Howes' team. Robertson Stephens' Juarez is "cautiously optimistic" about Loudcloud's future. "The opportunity is there, but no one is overly optimistic," he says. "There are just too many unknowns."

Whatever happens, Loudcloud is about to reveal what is perhaps the most interesting lesson of Netscape, and that is a lesson about Marc Andreessen: who he really is and where he wants to go. "I am not risk-averse, but I don't really like risk," says the man who inspired a generation to take risks their parents never dreamed of. "I don't like mountain climbing or snowboarding or hang gliding. Some people are excited about stuff like that, but I prefer to go home and relax. Risk freaks me out, and I try to minimize it. So while I enjoy the part of this venture that is a startup - the adrenaline that comes with the creation of a new company - it is less risky, since we have done it before. There is much that is known now that was unknown five or six years ago when we started Netscape. It is a better time to do a startup now."

The chair of Loudcloud is, in the end, not inclined to be a barefoot visionary. This is not a man who wants to leap, à la Jim Clark, from one new new thing to the next. "Do I want to put a lot of effort into something that is going to be around for a short period of time?" he asks. "If you do that, you just have to do it over again. Both here and at Netscape, it got more fun as time went on. It's even more fun now than it was three months ago. At the beginning, it's very frustrating not being able to execute. As time goes by, the more people we hire, the better the organization runs, the more customers we have, the more management we have, and so on - the more fun. Jim and others have a completely different outlook. For them, it's less fun as the company gets more established. But I am interested in staying on for the long haul, building something great and maintaining it."

Marc Andreessen showed the Valley just how golden the future could be, and in the process became the inspiration of every ambitious, techno-savvy kid in America. Nice work if you can get it, but what mattered more to him was the company he lost. After wandering in the wilderness at AOL, he returned to a hero's welcome, drawing the smartest minds in technology to his side. And what is it he wants to build? A business. A good, safe, solid, long-lasting business - and if that sounds boring, tell it to Andreessen. He'd be happy to hear it.