Guild Companies--Sun Talks Up Solaris for X86, Misses Big Opportunity

Metaphors are fun, so here's one: Sun Microsystems was early to the X86 party, got a little drunk on its own booze, left early, fell into a ditch after antagonizing its friends, got a bad hangover, and now it seems to have realized what happened, has choked down some aspirin and cleaned itself up, and come back to the party. Another sharp wiseguy came to the party while Sun was being belligerent--one might call that wiseguy Linux--and now Sun is trying to win back those X86 friends and influence people.

For 90 percent of the world, X86 iron does the job, and that is not going to change. Over the course of the past year, Sun has thankfully figured that out. But the operating system that runs on that X86 iron is very much up for grabs, and Sun very much wants to win a much larger share of installs on X86 iron than it has been getting in recent years. That's why Sun has been waving the Solaris on X86 flag for the past few months.

Last week, Sun announced that it has more than 300,000 registered licensees for Solaris 9 for X86. This may be a small slice of the 4 million or so new servers that the world's companies buy each year and the untold number of existing servers that change operating systems every year (which is probably a few million, because of the wide variety of operating systems in the X86 server market). But it is a big number for Sun, which needs millions of Solaris for X86 users if it is to beat Windows and Linux.

Sun delivered Solaris 9 on 32-bit Intel chips back in February after pulling it from the market with Solaris 8 and threatening not to put Solaris 9 on X86 back in May 2002. Sun was smart to put Solaris 9 for X86 back into the market, and should be commended for making the right decision after making the wrong decision to remove it to try to protect its Sparc/Solaris business. The people at Sun who think they can offer cheaper boxes than Intel iron to do the same work will probably turn out to be wrong in the long run. The volumes are with Intel, not with Sparc, no matter how many threads are on a chip.

The question is, can Sun make any money selling Solaris for X86? It's hard to say. Solaris 9 for X86 is free for non-commercial uses, as was the case with Solaris 8 on Intel Pentium and AMD Athlon processors. It costs $99 on single processor workstations and servers, and $250 on dual-processor machines. Prices rise from there. If every one of those Solaris licenses were paid for at list price and the average machine had 1.5 processors (Solaris is running on plenty of four-way and some eight-way machines), those 300,000 Solaris 9 for X86 licenses would have added up to a mere $52.35 million. It is fair to assume that a portion of the 300,000 Solaris for X86 licenses were freebies, although it is hard to say how many.

The point is, it is very difficult for Sun to make a lot of money selling Solaris 9. But it could significantly boost its installed base by giving Solaris for X86 away completely and then positioning it with fee-based services and support. Red Hat and SuSE do this with Linux, but you have to pay upfront to get into the game. If Sun just gave away Solaris 9 for X86 and let people realize they really do need services--this would be a big gamble, of course--it might be able to get hundreds of thousands of customers more to opt for Solaris X86 and then get a percentage of them to pay for support and services. Remember, the volumes that SuSE and Red Hat get for their licenses are not very high. Red Hat, for instance, was ecstatic because it added 2,600 new licensees for its enterprise server operating systems in the most recent quarter. It doesn't take much to start making money.

What Sun really needs is to stop obsessing with Sparc designs that (while elegant and technically sound) are years away and cannot help Sun now, and may not in the future. One option for Sun is to stop messing around and port Solaris to AMD Opterons, and to create a killer, cheap X86 server line all the way up to eight-ways with Solaris on it and just let the chips (including Sparc and Pentium) fall where they may. Sun has put out a call for Opteron experts (four software engineers and a manager, to be precise, according to its human resources Web site), but it had better already have a serious skunkworks in operation. Sun's software czar, Jonathan Schwartz, said this week that Sun would deliver Solaris on AMD Opteron processors, but Sun has to do more than this. It cannot be passive about Opteron. It has top be proactive, active, reactive, and postactive (if there is such a thing).

Sun's X86 iron, no matter what chip it is based on (Intel, AMD, VIA, whatever), has to be cheaper and offer a better value proposition than Linux on Dell PowerEdge, Hewlett-Packard ProLiant, or IBM xSeries servers running 32-bit Xeon chips. By partnering tightly and exuberantly with AMD, it might be able to shoot the gap between Xeon and Itanium and clean up with Solaris, which is a very well regarded operating system platform with the largest Unix installed base in the world. And with SCO ticking off potential Unix and Linux customers, Solaris on X86 could really clean up. And here's the real benefit: Sun could also support Linux and--dare I say it where the only founding member of Sun still at the company, chairman and CEO Scott McNealy, can hear it?--even Windows on its AMD midrange servers. Sun could make an agnostic AMD platform and then prove Solaris is better than Windows or Linux, and if it can't make that case, at least still sell some iron against IBM, HP, and Dell.

And if Sun doesn't have the stomach to take this shot, well, then maybe someone else should.