Elizabeth Corcoran, 05.27.02

How Sun hopes to innovate its way out of trouble.

Uh-oh--Javaman looks like he's in trouble.

Could this finally be the end? Like a comic book hero, Scott McNealy, chairman of Sun Microsystems, has thrilled the computer industry with his adventures for 20 years. The man who brought us the Java programming language and high-performance computers by the thousands relishes every bit of controversy. "If you all thought we were doing the right thing, then everyone would be doing it and we'd have no chance at making money," McNealy chided a crowd of investment analysts recently.

The latest episodes have been real cliffhangers, including $800 million in losses during the past 12 months and the departure of five lieutenants--including President and 15-year Sun (nasdaq: SUNW - news - people ) veteran Edward Zander. News of his exit took Sun's stock down 15% in a day, to below $7 a share, a low not seen since 1998. "Sounds like a buying opportunity to me," declares McNealy. Even so, Sun has no plans to up its stock buyback program, which spends $100 million per quarter of the company's $6 billion in cash.

McNealy says Sun will be profitable again by the end of the June quarter, thanks to a lineup of brand-new products. But Sun's problems may be far more fundamental than the current economic dip. Technology is turning against Javaman. For years the distant threat has been that the Wintel duopoly--Intel chips running Microsoft software--would defeat Sun from below. Until now Sun has been a wily player, inventing new technology that charmed customers into paying for Sun's unique features. Now the threat is not so distant.

Teams of small, cheap computers can now do just about all the jobs that once were the domain of Sun's machines. Customers who worried about Windows' porous security or tendency to crash are discovering that Windows is getting better--and in the meantime can turn to the hardy and basically free Linux operating system. IBM has also roared back, tempting customers with a smorgasbord of computer power and consulting services. In the nine months from July 2001 through March 2002, Sun's gross margins were 39%. Two years ago they were 52%. "Sun has historically been the leading Unix-based server in the market," says Rebecca Runkle, an executive director with Morgan Stanley. "But we are definitely seeing an inflection point as Linux and Windows gather steam."

Average selling prices have been dropping sharply. For instance, Runkle calculates that in 1997 Sun's "entry" level systems were about $27,000. By 2001 they had dropped to $13,000. IBM's cheapest systems sold for $7,200 in 2001. By contrast, Runkle puts Dell (nasdaq: DELL - news - people )'s average (for all its systems) at about $4,800.

Lots of current Sun customers still love the company. Sun kept ILX Systems, the division of Thomson Corp. that provides workstations to 80% of NYSE floor traders, without interruption through the Sept. 11 aftermath. Bernard A. Weinstein, ILX's president, says that the Sun system's ability to run reliably is invaluable. Similarly, Corporate Express in Broomfield, Colo. has systems handling $5 billion in office-supply orders every year. In 1999 the company began consolidating work from 43 Hewlett-Packard servers in different locations onto two of Sun's most powerful servers--and saved $10,000 a day as a result.

Other customers are less sanguine. Credit Suisse First Boston recently moved its trading floor software from Sun servers to Lintel systems (Linux on Intel chips). Morgan Stanley also recently moved off Sun products to a mixture of Lintel and Wintel systems.

Jeffrey Guilfoyle, a vice president with the two-year-old Solutionary in Omaha, Nebr., is building a center that will manage electronic security and firewalls for customers. Both IBM and Sun had attractive servers, he says, but both wrapped their sales into larger packages with additional licensing requirements and storage systems. Guilfoyle just wanted machines, so he went with clusters of Dell servers running Linux.

Royal Dutch/Shell's U.S. arm is consolidating all its database operations at three geographic hubs. "We're trying to drive the total cost of ownership down-considerably," says Richard Albritton, who manages the megacenter program. His pick: IBM.

So-called blades--powerful and thin computer boards that slide into a rack--are the hippest concept in computer centers. One refrigerator-size rack might hold 200 blades, providing tremendous computing power per square foot of floor space. IBM, Dell and Sun are rolling out blades this year. Price-performance data are still scarce, but the systems promise to deliver more computing per square foot in data centers at low prices.

"It's hard to be a major player and have your only capability be innovation," says Kevin Rollins, president of Dell. "It's about innovation, cost and quality. You need all three vectors going simultaneously."

Rollins' point is somewhat self-serving. Dell spends a spare 1.6% of sales on research, compared with 11% for Sun and 6% for IBM. But customers are beginning to agree with him. "We were a Sun-Solaris shop for five years and were happy with the platform but not with the cost/performance," says James Calloway, executive vice president of Nando Media in Raleigh, N.C., which runs Web sites for the McClatchy newspaper chain. These days Nando Media serves up 30 million to 40 million page views a month using 100 Dell servers, cutting his expenses in half.

Even executives at data processing outsourcer EDS, historically a Sun ally but now also a Dell fan, are circumspect when it comes to Sun's prices. "Anytime a customer starts to identify solutions as expensive, it's time to recalibrate the cost equation," says John Wilkerson, president of EDS' global alliances.

To this, McNealy retorts: "We couldn't be better positioned."

For the past 20 years, since the day it shipped its first computer ready to be plugged into a network, Sun has been a technology leader. Early Web builders frequently preferred Unix-flavored operating systems such as Solaris (and now Linux) to Microsoft software. Sun's Java language was a potent marketing tool. Between 1997 and 2000, Sun sales more than doubled to $18.3 billion, and net income climbed 2.5 times to a peak of $1.9 billion in fiscal 2000. Even as tech spending cooled, Sun executives kept a defiant stance. In late 2000 Zander told FORBES: "I think there's going to be a shift in capital spending, and it's going to be lucky for us." Companies, he said, "will keep investing in Internet infrastructure ... because that's going to let you gain global advantage and be more competitive."

Not quite. The drought in capital spending hurt Sun as much as any box manufacturer.

Sun executives have a comeback line. McNealy says his hottest product is its Sun Fire V880, a two-processor Ultra Sparc Solaris system that starts at $30,000, introduced last October. Sun is also filling out its low-end product lines. In February it dramatically broadened its support for Linux. Courtesy of an acquisition completed in 2000, Sun is also selling its Cobalt line of servers, which run Linux on Intel processors and now start as low as $1,000. Sun also rebranded--and repriced--an existing Ultra Sparc-based server that runs Solaris to sell for $995, calling it "$3,000 of value."

Even its longtime disdain for Intel technology is taking a backseat. In January Sun indicated that it was withdrawing support for running Solaris on Intel's chips. Customers were furious. "We were surprised" at the reaction, says Jonathan Schwartz, recently promoted to head Sun's software division. "Stay tuned," he says, implying that Sun will reembrace Intel.

Despite public posturing about Wintel's never becoming a viable rival, Sun executives have been debating for more than two years what would happen to their company when small servers became commodities. A skunk works panel of engineers has been working on lacing together small numbers of cheap computers to deliver the punch of an expensive Sun workstation. Last August this team took their conclusions to top execs: "If you keep building the same thing, you'll be fighting it out on narrower margins." Instead, they proposed a run for the high ground: Develop both software and hardware technologies that would help customers better manage all the diverse components in their data centers. It would be much like what Sun did when it was founded: taking commodity chips, software and networking gear and packaging them together for customers. Now, in a future strategy it is calling, for the moment, N1, it would take commodity servers, routers, storage gear and so on, and package them. "We're scaling up-going into a modern data center, finding thousands of components, and we're building the next system out of them," says Greg Papadopoulos, McNealy's chief technology officer.

But making good on that now-fuzzy promise is a daunting task. Papadopoulos says that it will take Sun another year to get its latest ideas about network management included in software, and a year or two beyond that to build specially designed microprocessors that will run the hardware in such systems.

The savviest, and most honest, executives at Sun realize that betting on technology shifts is a tremendous gamble. "If you look at all the major discontinuities in the business," said Zander in an interview in May, "it took several years to figure out what was going on. It's like being in a hurricane. Is the future about optical computing? Is it about N1? And then you have to go through several versions to get the products right."

Schwartz, who is now among the dozen managers who report to McNealy, says he thinks Sun works best when cornered. "We lost our way when there was no top dog to topple. We are much better defined by having to innovate through [problems] than by a marketplace that just douses the industry with free capital and says go buy more servers. So I think we're beginning to hit a stride again." We'll see.