Scott McNealy believes the best preparation for business is to play hockey. The teamwork, the individual risk taking, not to mention the bone-jarring collisions, are the closest analogy he knows to what is needed to function as the CEO of a technology company. In other words, he could never have taken Sun Microsystems in the direction it needed to go: Sun needed to become a software company.
Software is amorphous, invisible, full of promise that may or may not be fulfilled. Used car salesmen, with their goods at least visible to the naked eye, are paragons of virtue compared to software salesmen. How could Sun, a young and innovative hardware firm, make the switch to this tricky to manage, less tangible entity, a software company? In a word, it couldn't or at least it didn't, which doomed Sun.
And why couldn't it? Go back to exhibit A, Scott McNealy. He thinks the best training for business is hockey. The intangibles of software, where a human presence and commitment has to substitute for tangible goods, where an ability to think abstractly about a roadmap, shared with customers through a finely calibrated sense of mile markers, has to substitute for discrete products -- how likely are these subtleties to come from a hockey player? That's doubtless an insult to many hockey players. But it's still fair to say about Sun.
A software company is a very different thing from a hardware company. Sun developed excellent software in NFS, Solaris, Java, and many other examples. It had a thriving software engineering force. But that software talent was always dominated by and made to serve the purposes of its hardware rulers.
A hardware maker has to compete head to head, beat the competition on defined metrics and produce a reliable product on a definite schedule. When Sun's software culture produced Solaris and Java, it gave a hardware company an unusual opportunity to launch a software business with assets that were valuable if less precisely defined. A more mature Sun might have understood this and become a two-business entity, still thriving today. Instead Sun lost 80% of its value between November 2007 and November 2008, and was swallowed up in an April 2009 fire sale to Oracle. With both Java and Solaris, the hardware culture appropriated the initiatives of the software culture, to Sun's detriment.
McNealy came as close as I've ever seen to recognizing this in his talk Feb. 24th at the Churchill Club in Santa Clara, Calif. He said of Solaris: "The mistake we made was putting it on our own hardware. If we hadn't metal-wrapped it, it would have been more widely adopted. If we had put Solaris early on an Intel box, Linux never would have never happened."
To do what's described above at the right time would have required an assertive software leadership inside Sun, someone willing to stand up to the hardware culture and say Solaris was going to run on multiple chip architectures, including Intel's, because that's where Solaris' opportunity lay. For any Sun executive to take such a stand, however, meant he would incur the wrath of powers greater than himself and he would hang by a thread, dependent on the inclinations of his CEO.
McNealy, the one figure who could have given the nod to such an initiative, was himself imbued with the hardware culture, where you collided with your competitors, overwhelmed them with your strength and bested them at the goal. What the software side of Sun was saying was that it needed to cooperate with competitors, make Solaris more understandable to the outside world and more portable to other architectures. Without empowered software leaders, and with the CEO having no vision of what a software company must be, how was that to happen?
Sun eventually produced Solaris for x86 in order to claim that Solaris was a multi-chip system, but for several years, it didn't take the port seriously. Solaris initially ran like a snail on Intel, and few outside of Sun took it seriously. Then Linux became a serious threat, eroding Sun's low-end Unix customer base, and Sun revamped its offering into a suitable competitor. But it was too late.
To illustrate what a slow learner Sun was on this point, the same issue arose with Java a few years later. Java came along as customer interest was peaking in the Internet. Many corporate programmers had no idea how Java would help them thrive in the Internet age, but it was clear things were changing and here was the first modern language that seemed in step with the times. In fact, a whole battery of languages would spring up to serve the new era, but as a viable newcomer, Java had no peer as it gained traction inside companies in the mid-1990s. If the "metal wrapping" of Solaris on Sparc had been a mistake, Java offered an opportunity recoup to reach out to legions of new customers with vendor-neutral software.
"When we go into a company, we say, "Java, Java Java,' then sell servers," a prominent Sun executive told me at JavaOne in 1999 as the dotcom boom was well underway. Java was a passport through corporate resistance and a trump card in a hardware/software sale. If the product came from the inventor of Java, it had to be right for the times, and Sun joined in the boom.
In the midst of this frenzy, the Sun hardware culture still secularized the universal appeal of Java to its own hardware. You could write code once and run it anywhere, thanks to the portability of the Java virtual machine, but if you wanted the JVM to run where it would perform the best, then you needed Sun hardware, was the concluding message at numerous JavaOne speeches and session presentations in the dotcom boom. This was eventually revised, but the proprietary air Sun maintained over Java left its mark.
Sun had reorganized and created within its ranks JavaSoft as an independent business unit. JavaSoft executives in effect headed a software company that wanted to see the language as widely established as possible and they welcomed the support of IBM, HP, and other large companies which, in many other areas, were tough competitors. But JavaSoft's message was mixed. It extended an olive branch with one hand, even as it affirmed it planned to keep a competitive advantage over how the Java Virtual Machine ran under Solaris. Sun needed to neutralize the tensions, but McNealy, with his love of showmanship, tended to increase them. HP was an excellent "printer company," he would say, while IBM excelled mainly at charging you for services; only Sun coud truly offer Java performance.
Java's success seemed only to make McNealy more strident, as Sun realized how much Java had come to mean to some of its competitors and how much they wished it could be taken out of Sun's hands (and put under a public standards body).
Dozens of young software companies sprang up to capitalize on the portable new language, and Sun struggled to cope with their demands. On the one hand, it found them troubling in their insistence that Sun not compete directly with their young businesses. On the other, they were a help in establishing an ecosystem around Java and counterbalancing the weight of HP, IBM, and other big Java adopters.
JavaSoft as a software company needed to pick its spots, compete on tools and let the chips fall where they may. It didn't do that. Nor at times did it know whether the Java ecosystem consisted of helpers or competitors. Sun started out strongly in Java tools, a legitimate endeavor for its software culture, then backed away, as it caught flack that it had an unfair position as the language's caretaker. The hardware culture was unpracticed at playing the role of honest broker.
After establishing the Java Community Process, where the votes of many small companies -- Instantiations, Bluestone, BEA Systems, Borland, JBoss -- could offset IBM, Oracle, and HP, Sun realized it had missed a major opportunity and waded back into tools, a burgeoning field. It tried to catch up through acquisition, spending hundreds of millions in 1998-99 to acquire NetBeans and Forte Software and invest in supporting software from Netscape, i-Planet, and SeeBeyond, money that would soon be sorely missed.
As Sun re-entered the tools business, IBM issued as open source code its internal environment for sharing Java tools. Sun wanted to be the central authority, the pivot point, for Java tools. Instead, through the IBM programmer's workbench, many third-party tools started using its integrated development environment for their own operations and to share files with other third-party Java tools.
IBM denied it sought to impact Sun's business, but the tensions between the two were palpable over what IBM claimed was Sun's reluctance to live up to a pledge to create an international Java standard. IBM customers depended on Java, but Java was still under Sun's proprietary control. The Eclipse workbench was IBM's response. It blanked out Sun's second attempt at a profitable tools business, in part by capitalizing on the open source code enthusiasm among Java programmers. If it had really been aimed at Microsoft, as IBM later claimed, it would have been called Storm Cloud Over Visual Studio, not Eclipse.
Just as in tools, Sun backed away from, then re-entered the production of key Java middleware, long after others were already established in the field. By the time McNealy relinquished the CEO's keys to Jonathan Schwartz, the jig was up, the hardware culture had exhausted its options and in desperation it was willing to try something else. Once again, Sun was reacting to market conditions too late, with too little business discipline, too little understanding of what its software engineers would have done if left to their own initiative.
Schwartz, the new CEO, was not part of the hardware culture and had the makings of a fine software company executive, provided he had been appointed seven years earlier and been given real power upon assumption of office. Instead, he had been given a badly played out hand. He and McNealy, his mentor hero, clashed in the final denouement. Schwartz understood that IBM would be a better repository of the Java trust than Oracle and favored a sale to IBM. McNealy, always resentful of IBM, insisted on a sale to a personal friend with a seemingly more palatable deal, Larry Ellison.
The true proportions of this Titanic mis-pilotage were missing at the Churchill Club event on the evening of Feb. 24. "I'm still not satisfied with the discussion," said Larry Hammond, vice president of Morgan Stanley and previously an analyst with Credit Suisse First Boston, in the last question of the evening. "Why did HP, IBM, Dell, and all other major companies recover after the bubble burst? The only major one that did not was Sun. Since then, we've seen data center servers, departmental servers, database servers, ecommerce grow like crazy. How did Sun miss the whole thing?"
It was the most pertinent question of the evening, one that called to account the entire misguided hardware culture. IBM and HP, it might be noted, are now both hardware and software companies, not one or the other.
As McNealy stayed silent, Ed Zander, former Sun president and emcee of the occasion, answered true to form: "Larry, I hear ya. It was a blast. I was coming up 101 tonight and I would have loved to see those buildings still showing the Sun logo. But it's time all of us moved on. We all loved the Sun experience but it's time to move on. If it's all right, I don't want to go there tonight."
The avoidance of difficult questions, the unwillingness to make a painful call -- how much farther do we need to look to understand why Sun failed?
Charles Babcock is an editor-at-large for InformationWeek.