But when Andreas Bechtolsheim, Sun's first employee and the creative force behind many of the Silicon Valley giant's early workstation computers, joined McNealy on a stage in San Francisco in early 2004 and announced he was returning to the fold after nine years away, McNealy gave him a big, happy hug.
"It's one of the most exciting days I've had in a long, long time--in 22 years of Sun Microsystems," an effusive McNealy declared.
Indeed, it was a defining moment in the company's five-year effort to recover the prestige and financial health it lost when the dot-com bubble burst. While it's way too soon to say Sun is back on track, the return of Bechtolsheim, aggressive improvements in products and a healthy dose of humility among Sun's executives mean the troubled company and its investors have more cause for optimism than they've had in years.
Sun is in the midst of an effort to recover the prestige and financial health it lost when the dot-com bubble burst.
In a two-part series, CNET News.com takes a look at the new products and plans that could get Sun back on track, and explains why some other areas may not be so easy to fix.
Sun is hardly out of the woods, but McNealy and outspoken President Jonathan Schwartz may finally have a map that could lead them into the clear.
Sun engineers are creating two new lines of servers almost from scratch, while a radical new family of the company's own Sparc processors could add new interest in the Silicon Valley giant's traditional server line. If that's not what customers want, Sun has started selling its first serious machines based on mainstream x86 chips--with Bechtolsheim leading the project.
On the software front--long a weak spot for the company that invented Java software technology--Sun has begun retooling for a world where open-source software is common. Instead of fighting a trend that has deeply hurt sales of Sun's own Solaris operating system, McNealy & Co. are finally embracing it by offering an open-source version of their flagship operating system and promising all its other software will follow suit.
Sun is also building alliances with rivals such as IBM and Microsoft and spending billions of dollars on aggressive--if risky--acquisitions such as that of storage equipment maker Storage Technology.
At the very least, Sun has managed to generate a little positive buzz thanks to a partnership announced two weeks ago with Google, which is run by another Sun alumnus, Eric Schmidt. Though the announcement was big on concepts and short on details, Wall Street loved it. Sun's stock jumped 7 percent the day the deal was announced.
To truly start growing again, however, Sun needs to generate a lot more than buzz. It still must turn around market share losses, convince customers its servers aren't overpriced relics and find takers for server software that hasn't been terribly popular. And some initiatives, such as its OpenSolaris project and its
All the while, it will face near-constant price pressure in its core business selling servers.
A good turn?
There's a mood change, ever so slow, starting to take place around Sun. In a 10-author report released in September, the influential analyst firm Gartner upgraded Sun from "caution" to "promising" based on financial improvements, strong technology, innovative pricing and good relationships with programmers.
If a turnaround is imminent, it's been a long time coming. At its peak, Sun revenue grew 60 percent year-over-year to more than $5 billion in the third quarter of 2000, while net income grew 85 percent to $510 million.
But Sun was slow to respond when the tech bust stripped it of many customers. On top of that, new technology trends such as x86 processors and Linux overtook Sun's once-glamorous product lines. Sun's product quality slipped, McNealy has acknowledged. Revenue plunged, profitability evaporated. More than 13,000 have been cut through five large layoffs since 2001.
In its most recent quarter, ended June 30, Sun revenue slid 4 percent to $2.98 billion--that's 41 percent from its peak--while the company edged above break-even in profitability with net income of $50 million.
At the same time, Sun has lost share in its most important market--servers running the Unix operating system. In the second quarter of 2005, Sun's Unix server sales dropped 7.4 percent to $1.38 billion, while second-place Hewlett-Packard grew 6.3 percent to $1.22 billion and hard-charging IBM grew 32.8 percent to $1.18 billion.
Critics say Sun has flailed about for a strong product road map since the wheels came off five years ago. The question now is whether the latest iteration of that map will cut it.
Sun eased into the x86 market in 2002. That effort went up a notch in September, on the release of the "Galaxy" designs that use Advanced Micro Devices' Opteron processor. Development of the servers began at Bechtolsheim's start-up, Kealia, which Sun acquired.
The dual-processor systems cost about $5,000 for a midrange configuration. Sun prefers customers use Solaris x86 as the operating system, but also sells them with Linux from Red Hat and Novell and supports use of Microsoft Windows.
So far, Gartner said Sun has climbed from nowhere to the sixth-biggest seller of x86 servers in the second quarter of 2005, when one out of every five Sun servers shipped were x86 machines. Sun executives hope Galaxy will carry it to fourth place by the end of 2006, which would top NEC and Fujitsu.
Merrill Lynch analyst Richard Farmer has forecast that Galaxy will add more than $150 million new revenue for Sun's fiscal 2006, which ends June 30, 2006, and more than $350 million in fiscal 2007.
The next step is harder: Market leaders HP, IBM and Dell beat Sun's x86 server revenue by factors of 19, 12 and nine, respectively. It could be hard to convince customers to switch. "We're kind of standardized on Dell. Why bring on another hardware platform?" asked Chris Hughes, director of information technology at regional carrier Independence Air. "It would just raise our costs."
Galaxy has other big challenges, starting with marketing. Allen Smith, chief information officer at accounting firm Virchow Krause & Co., is responsible for management of about 100 x86 servers, but he hadn't yet heard of Sun's x86 move until informed by a reporter. Even among Sun's customers, "the dominant majority don't even know (about Galaxy) yet," Sun's Schwartz acknowledged.
Sun has to convince customers the Galaxies are worth the price. Smith, for example, said he wouldn't consider Galaxy servers unless Sun improves its Microsoft support to the same level that HP, Dell and IBM offer--or offer a 50 percent lower price.
Also, there are still barriers between the x86 servers and the traditional Sparc-based products.
"Almost all my infrastructure is Sparc-Solaris based," said Roger Smith, senior systems administrator at the Mississippi State University's engineering research center. "I still am not totally sure I understand where x86 Solaris fits in the world," because the software for the Sparc version can't simply be moved to the x86 version. "From where I sit, it's a whole different operating system."
The Galaxy strategy extends to eight-processor machines due in the next few months into the relatively high end of the x86 market, from which Dell and HP have withdrawn. There, Sun is banking on its operating system advantage.
"I think the market is going to really rapidly grow," argued John Fowler, executive vice president of the high-end x86 business. "There's no real operating system for it" besides Solaris.
A second key plank in Sun's turnaround strategy is the rejuvenation of the Sparc processor family. Sparc servers remain Sun's most important business, and boosting Sparc sales is probably the single easiest way for the company to restore financial health, even though market researcher IDC forecasts the $19.1 billion market for Unix server servers will shrink by $200 million during the next four years, while Linux and Windows sales continue to grow.
"No matter how Galaxy takes off in the marketplace...it's going to be off a comparatively small base," Schwartz said. "We're in the billions of dollars of opportunity in the Solaris-Sparc marketplace." Sparc server sales also tend to tow along other business, such as data storage and customer services, Sanford C. Bernstein analyst Toni Sacconaghi said.
Here's the Sparc lineup today: UltraSparc IV, released in 2004, and its new IV+ sequel released this month. Both feature a dual-core design, which puts two processing engines on the same slice of silicon.
Next in the Sparc rejuvenation plan comes Niagara, an eight-core processor that can execute 32 simultaneous instruction sequences called threads while consuming only 56 watts of power--less than half of an Intel Xeon, which can handle handle four threads.
There are indications that Sun has already piqued some customer interest. Internet auction giant eBay is among the customers trying the machines, using them in its PayPal online payment division, sources familiar with the testing program said. eBay declined to comment.
Not surprisingly, competitors are quick to downplay Niagara. "For Sun's declining installed base, it's going to be a great product," Pat Gelsinger, general manager of Intel's Digital Enterprise Group, said.
Maybe so, but it's a customer base Sun certainly needs to keep happy, and those high-end customers can be leery of x86 machines. "What we need is something that is solid or stable," said Carsten Larsen, general manager of commercial development for Australian utility provider ActewAGL.
And Sun has other plans. Next up is the "Advanced Product Line" partnership to sell servers with Fujitsu's Sparc64 VI "Olympus" processor beginning in late 2006.
Then comes Niagara II, built with a more advanced manufacturing process. Its features include hardware acceleration of at least seven cryptography algorithms, the ability to cooperate with Solaris to classify network traffic and send it to the appropriate processor core, and a built-in 10 gigabit-per-second Ethernet interface, said David Yen, executive vice president of Sun's Sparc group. In addition, it will be possible to make systems with multiple Niagara II chips.
Sun's public chip plan extends as far as "Rock." Where Niagara is geared for network-oriented tasks such as application servers or Web site hosting, the Rock chip family due in 2008 is designed for back-end tasks such as databases, where a single thread must execute as fast as possible.
Rock also will accelerate Java programs, facilitating the "garbage collection" process by which unused memory is freed for use, Yen said. Initial hardware design for both Niagara II and Rock will probably be completed in the first half of 2006, Yen said.
Rock, Niagara and Galaxy aren't a sure bet, of course. But many customers still would double down on Sun.
"Sun has been around for a long time," ActewAGL's Larsen said. "We're confident they're going to be around for some time to come."
Tuesday: Sun can put a great shine on its chips and servers, but will it ever be able to fix its disappointing software and storage equipment businesses?