By Paul Sloan

(Business 2.0) – Marc Andreessen is scribbling so furiously that he's about to tear through the paper into the table below. He whips off a few oblong shapes. "Somewhere in here is where we are now," Andreessen says, adding circles upon circles to illustrate the state of the high-tech industry. He jabs his fine-point felt-tip into the table to show the peak of the bubble. "CyberCash--remember that piece of shit?!" CyberCash set out to replace cash with digital currency and became a dotcom darling. "Because it was doing what?!"

Andreessen's voice is rising in volume and pitch--he sounds like a cassette tape running too fast. Now he's waving his arms above his 6-foot-4 frame, both middle fingers outstretched toward the ceiling. "People were saying CyberCash was essential to the Internet! Why? I was like, 'Screw you!'" Some fellow diners in the Palo Alto restaurant where Andreessen is holding forth peek warily at this big guy who's all worked up. "I shorted that stock all the way from 50 to zero!" Andreessen roars, interlocking his fingers, cracking his knuckles, and flashing a big smile.

This is Marc Andreessen after a couple of Cokes and some dim sum: enthusiastic, tireless, excitable. He's disdainful of the dotcommers who gave the tech industry and Silicon Valley a bad name; he's glad they all went home. Now the Valley is once again safe for serious people with legitimate business plans who will create the next wave of vital, world-rattling technologies.

People, as he views it, such as himself.

To understand Andreessen and what he is up to these days, understand first that he sold Netscape to AOL for $10 billion and thinks he failed. Sure, he co-created the first hit browser, Mosaic; and Netscape, which he co-founded, became an epoch-defining company. But Andreessen was never just the barefoot boy wonder of the Web of bubble-era legend. Today, at the ripe old age of 32, he is still trying to realize the dream that he has harbored since he lit out from Champaign, Ill., for Silicon Valley in 1994: to create a giant, enduring company.

Why? It's not the money--he walked away with $92 million after AOL bought Netscape. Andreessen wants to dispel a notion that has trailed him since Netscape's fade: namely, that he's one of the biggest one-hit wonders in business history. "Marc," says his friend and former Netscape executive Eric Hahn, "feels a personal need to show the world he's more than just the browser dude."

The little software firm over which he now presides, Opsware, is at first glance an unlikely pulpit from which to launch such a crusade. It has 105 employees, has never made an operating profit, and has revenues that wouldn't even reach the level of a rounding error at many of its rivals. Yet Andreessen says this company, like his first one, will change the way every business in the world runs its technology.

On a sunny spring afternoon, the browser dude is crammed inside a windowless conference room with a dozen Opsware executives when he blurts out, "Jayson Blair." Dead air ensues. It's a classic Andreessen moment. On the wall, a screen shot of a new Opsware product shows a real-time readout of the operations of a sample company's myriad servers--what software is running where, what time any changes were made, and, most important, who did what to which machine.

OK, so what does Jayson Blair, the ex-New York Times fabulist, have to do with Opsware? It's simple, Andreessen explains. Corporate IT organizations are full of technicians tinkering with the servers, and there's no way to reliably know who made what changes to which machines. One thing Opsware does is let the boss keep track of this, right down to the second. So if you happen to have a Jayson Blair in your IT organization--someone who, say, lies about a fix he made--Opsware will bust him. The customer doesn't have to assign a team of investigators to figure out what in the chain went wrong. "Every IT shop can have a Jayson Blair, and odds are everyone does," Andreessen says with a smile. "I want to reorient our story around Jayson Blair."

Put less metaphorically, Andreessen believes Opsware offers an antidote to the chaos of the modern corporate data center. The party-hearty spending on information technology unleashed by the Web boom has left companies awash in mismatched servers, balky software, and IT managers who are clueless about what's going on in their own shops. Today roughly half of all IT spending goes to pay people just to keep things running. Opsware's main function is to let IT administrators make changes to thousands of servers with the click of a mouse, helping to tame one of the most maddening and costly problems in business today.

Once again, Andreessen is going at a piece of paper, this time scratching out arrows linking a computer terminal with a database and hundreds of servers. He's illustrating what Opsware calls its "virtual model"--what makes it unique--and jokingly observing that the whole thing "gets pretty religious pretty quick." The core engine of the model is called "the way"; another part is called "the truth." Built into "the way" are "knowledge modules" that know everything about the 70 or so products that Opsware supports--servers and software by IBM, Microsoft, Oracle, Siebel, Sun, you name it--plus whatever is added for a specific customer.

The "model" lets IT managers simulate what's happening on all their servers so they can do tests and then make sweeping changes. The model alerts you if someone makes a change to a server. Then it will ask you if you want to change your other servers to match that one, or undo the first change that was made. And it can do this across multiple data centers spread across the globe.

When a hacker unleashed the Slammer worm on Microsoft's SQL Server last January, for example, the bug froze thousands of corporate websites, shut down ATMs, and slowed the Internet itself. The attack didn't have to create such havoc. Microsoft had sent out a security patch that would have killed the worm, but many administrators had simply failed to install it.

"It didn't hit people's radar until it was too late," says Steve Lapekas, head of Web hosting services for EDS, a big Opsware client. Part of the problem is that troubleshooting and installing a patch on a single server can take several hours--which is why so many companies were unprepared for the more recent attacks by Blaster and other bugs. With Opsware, Lapekas says, he quickly determined that 150 of his servers needed the Slammer patch. "Then," he says, "it was just, bam, and it was done." Without Opsware, he figures the process would've taken weeks--and the worm's effects would've been disastrous.

That Andreessen is even around to muse about corporate Jayson Blairs and "the truth" is something of a miracle. The back story is part of high-tech lore: He invented a graphical way to navigate the Net and hooked up with Jim Clark, and together they took on a losing battle with Microsoft. After AOL gobbled up the company, for the first time in his life the hyperenergetic Andreessen had a hard time getting out of bed. To work off his aggression, he began lifting weights and took up kickboxing. He also became AOL's chief technology officer. At AOL headquarters Andreessen would sometimes retreat to his office and crank up the stereo. "It wasn't part of my job to change anything," he says. "There was really just nothing to do."

There was, however, the burning urge to come up with the next big idea, and in September 1999, he and a few old Netscape pals launched a startup called Loudcloud to run the websites of new Internet companies. The crew, led by Andreessen as chairman and his longtime colleague Ben Horowitz as CEO, developed software called Opsware to automate much of the process.

Thanks to the Andreessen pedigree, Loudcloud quickly raised $23 million. But inside the company, Andreessen soon began preaching that the VC money was going to dry up. He pushed his colleagues to raise a third round of financing even though it would mean giving more shares to outside investors. A number of high-level employees were so outraged that they demanded a meeting with board members.

"What the hell?" hollered Tim Howes, Loudcloud's CTO and a co-founder. "Why aren't we raising a large amount of debt?"

At the time, tapping into the debt markets was easy. Remember the telecoms?

"The risk is too great," Andreessen shot back. "We'll lose the company."

"Dude, that won't happen," Howes retorted.

In June 2000 the company took the venture money--$120 million--but the VCs still paid too much. By January 2001, Loudcloud had $80 million in cash and was burning through roughly $12 million a month. When Loudcloud went public in March 2001, investors bought shares at $6 apiece--two-thirds less than what some venture capitalists had paid nine months before. Loudcloud became the only Internet IPO of 2001, raising $163 million. "They raised a ton of capital at valuations that in retrospect are obviously hard to justify," says Roger McNamee, a partner with Integral Capital Partners, an investor since the early days. "They were smart. All their competitors went out of business almost immediately."

Still, most of Loudcloud's customers were dotcoms dying by the day. Andreessen took to the road and signed up blue-chip clients like Blockbuster, News Corp., and USA Today, but many big companies were wary of handing their IT operations over to a teetering startup. Several were interested in the software, however.

That sparked an Andreessen brainstorm: Loudcloud should gear up to sell the software separately. The shift, code-named Oxide, led to tumult inside the company. "All our Stanford MBAs were ready to revolt," Horowitz recalls. "They were saying, 'This is ridiculous. This company isn't focused.' But they didn't understand how bad things were getting."

In June 2002, Loudcloud sold its Web hosting business to EDS for $63.5 million in cash. The deal gave Loudcloud life, but it also meant 120 people would lose their jobs. Horowitz had to deliver the news, and for the sake of appearances, he barred his hard-to-manage partner from the office--sending Andreessen a mile down the road to the Sunnyvale Sheraton, where he sat taking phone calls from investors and the press. "Ben kept me just far enough away so that no one knew I was grinning," Andreessen acknowledges.

For EDS, the main attraction of the deal was the Opsware software itself, which EDS is paying an additional $52 million to license over three years. EDS is deploying Opsware on more than 50,000 of its own servers across 156 data centers; it says the software will save it $100 million a year in IT costs. Last September, Loudcloud shed its bubble-era name and became Opsware.

The company has since landed a dozen customers--some paying $500,000, others a few million--and a key partner: In June, Andreessen reached a deal under which Hewlett-Packard will sell Opsware along with its heavy-duty software for data centers. "Our customers kept bringing up Opsware," explains Nick van der Zweep, HP's director of utility computing. The alliance lets Opsware tap into HP's 200 utility-computing sales specialists. As Gartner analyst Ronni Colville puts it, "HP gets them a lot more feet on the street."

Andreessen's main mission now is to evangelize the benefits of automation to spending-wary CIOs. It's a tough sell, but companies everywhere have to do more with less. MetLife, for example, saw its IT budget fall by $60 million, to $900 million, in 2002. At the same time, the company has been moving more services onto the Web, with online activity doubling in the past two years.

So far, MetLife has installed Opsware on 500 of its 3,500 servers. Steve Bozzo, MetLife's CIO for international operations, says Opsware lets him complete projects in "minutes or hours instead of weeks or months." The company's IT staff of almost 3,900 is fortunate--MetLife hasn't let workers go during the downturn--but because of Opsware, Bozzo says, he's been able to move technicians out of the data centers to work on other projects. The software provides the "nirvana you look for in any new technology," Bozzo says--it improves operations and lowers costs.

That's exactly the message Andreessen wants to get across to corporate America, and one of his chief jobs these days is trying to figure out how best to deliver it. He says an important lesson he took away from his stint under then-AOL president and noted marketing whiz Bob Pittman is that selling any product demands a direct and simple pitch. Clever for the sake of clever is deadly.

So when two execs from Opsware's ad agency, Palo Alto-based NIA Creative, met with Andreessen in June to explain how they planned to interview Opsware execs to work up the ideal marketing message, Andreessen barely let them talk: "That's exactly what I don't want," he said. "I have no interest in being creative. I have no interest in being interesting. I just want 'Opsware automates your servers.'" In a follow-up phone call, Andreessen barked even more specific orders: "I don't want any ad copy over three syllables."

These are some of the last vestiges of Andreessen the enfant terrible of business. He has grown up a lot since his Netscape days, when he often blew off meetings. He's also matured enough to know he shouldn't manage people day-to-day. "Managing is an endless series of other people's personal problems," says Andreessen, who, at Netscape, once had 1,200 people under him. "People are complicated. They do stupid things. They are weird."

His distaste for managing means that, however much Andreessen wants to build a big company, he stands no chance without a partner like Horowitz. Among other things, Horowitz acts as interpreter for the many Jayson Blair-like moments of bewilderment that Andreessen creates. Horowitz says Andreessen is generally thinking too much about the future to be effective as a manager. "Marc's operating at this intergalactic level," says Horowitz, who, at 37, has worked alongside Andreessen for nine years. "To have him interacting too much with people on the ground wouldn't be a good idea. He needs translating."

Andreessen is trying to translate his thoughts about changing the world, but first he needs to pull Kettle out of the cords cascading from the racks of devices in the back room of his house. Kettle is his cat. "Hey!" Andreessen shouts. "You're shredding my LAN!" Andreessen's Palo Alto home is a gadget freak's dream: five computers and four TVs, all linked, including the 60-inch flat-panel plasma in the room off the kitchen. He just installed a Cisco router and virtual private network software so he can tap into his home network from his other house, in Hollywood, and download MP3s to his Jaguar convertible.

Changing the world is 99 percent about being in the right place at the right time, Andreessen is saying as he kicks back in a fat leather chair fit for a psychologist's office. That's what happened at Netscape, he says, "where it was sort of like, holy shit, this thing is taking off with or without us." Andreessen says he actually prefers building a company in a down economy (engineers aren't constantly jumping to startups) and working on big, corporate problems. (Netscape, he likes to point out, wasn't just a cultural phenomenon; its main business was selling software to corporations.)

But isn't all this invisible plumbing--like Opsware stuff--dull as dirt? It's certainly not going to land Andreessen on the cover of Time magazine, as Netscape's IPO did in 1995. "I don't want to be more famous," Andreessen says. "I'm doing what I like to do, which isn't the case for a lot of people."

Yet Andreessen is a guy who thinks big. He clearly enjoys the status, influence, and money that come with splashy business success. And how could any follow-up to Netscape, which, in effect, created the Internet as we know it, not be something of a letdown?

The answers may lie in the fact that Opsware, mundane though it may seem, is central to a grand and, for Andreessen, thrilling vision. He argues that his company is creating the foundation of tomorrow's world. He sees a time when pretty much everything--not just PDAs and cell phones, but shoes and baseballs and lightbulbs--will contain chips that are networked together by billions of servers running millions of pieces of software. That vision demands a fantastic level of automation--put another way, a tremendous amount of Opsware--or there won't be enough tech workers on the planet to maintain it all. Already, things are edging in that direction: Despite the IT-spending slump, servers continue to get cheaper and to proliferate; companies bought 1.3 million of them in the second quarter, up 18 percent from a year ago. "The big picture is completely positive," Andreessen says.

While Andreessen's tiny company has a head start, the competitive landscape is quickly changing, as companies such as IBM, Sun, and Microsoft try to automate the data center and offer what is variously called utility or on-demand computing. The idea is to let companies tap into computing power as they need it, creating a seamless, automated system. As the big guns try to make such fluid computing a reality, they have acquired or partnered with software startups rather than trying to reinvent the technology. IBM, for example, bought former Opsware rival ThinkDynamics in 2002.

Some analysts speculate that Opsware is destined to be swallowed too, possibly by current partner HP. Opsware is debt-free and cash-flow positive, and has $61 million in cash. Though its stock price is up 16-fold since the company shed its hosting business, to around $8--arguably high for a company that's yet to turn a profit--with a market cap of $650 million Opsware would still be affordable for any deep-pocketed player. "Marc truly believes Opsware can be independent," says Corey Ferengul, a Meta Group analyst. "Because if he's right about the way this plays out, this market is absolutely huge. But if things are going well and HP makes an offer, it might be too good to turn down."

Such an outcome would have eerie echoes of Netscape: Andreessen, who owns 13.9 percent of Opsware, would make a ton of money, but again would have failed to build a lasting corporate monument. Andreessen is adamant that he doesn't want to sell and dismisses those who suggest he might have to. "The analysts," he says, "are smoking crack." He claims he can compete with the likes of IBM because enough customers want software that doesn't favor a particular brand of hardware.

But even if Andreessen's vision plays out and he does create an independent company to rival those he often points to as models--Oracle, Dell, Intel--he still may have trouble shaking the browser-dude moniker. Opsware's market could end up dwarfing that of Netscape, whose best-known product, after all, was free. If that happens, he'll never be regarded as a one-hit wonder. But it's hard to imagine Opsware ever matching Netscape in historical importance or sheer mythic resonance. Andreessen contends he doesn't think about such things. Call him browser dude. Even one-hit wonder. Call him whatever you want.

"I just really don't care what people say about me," he says. Should Opsware fail or get bought out, Andreessen says, "I'll be disappointed, but I'll still be the same person the next morning." Which is another way of saying he'll get out of bed, pump some iron, and work like hell to try it all over again.

Paul Sloan (psloan@business2.com) is a senior writer at Business 2.0.