The curious case of Sun's hardware biz

Not being clear about the plan has its costs, as Sun's preliminary results for the fourth quarter of fiscal 2009 show, after they were announced earlier this week ahead of the Sun shareholder vote on the Oracle takeover. Sun said that based on a first pass on the numbers, sales in Q4 would be in the range of $2.58bn to $2.68bn, down between 29.1 to 31.7 per cent over the $3.78bn in sales in fiscal 2008's fourth quarter. Sun added that it would lose between 24 cents and 34 cents per share, which I estimate to be a loss of between $179m to $253m (provided Sun hasn't done a lot of share buybacks in Q4).

The amazing thing, given the economic meltdown and the considerable uncertainty about Sun's hardware and software product lines, is that Sun's sales in Q4 weren't off 40 or even 50 per cent compared to last year. And if they were, Oracle might have been under intense pressure to back out of the deal.

The fact is, Sun is a dot-com darling that was paired with the other big Silicon Valley startup - Oracle - to become the de factor standard for dot-com companies a decade ago. And that pairing did indeed make Sun and Oracle a heap of cash. In the fiscal 2000 year, you have to remember, Sun posted sales of $15.7bn and brought in $2.39bn in operating income and $1.85bn in net income. For that year - the best in Sun's history - the company's sales were growing at 33.2 per cent annually and net earnings were growing at a stunning 80 per cent.

Here we are, nine years later, and Sun's "Niagara" family of Sparc T servers are not competitive with the most recent x64 boxes. Its high-end Sparc Enterprise servers (rebadged boxes using Fujitsu's quad-core Sparc64-VII processors) have fallen behind IBM's Power6-AIX iron, and can barely keep pace with the Itanium-based Unix boxes from Hewlett-Packard, which are themselves looking pretty long in the tooth with so many delays in the "Tukwila" quad-core Itaniums.

Sun's "Galaxy" family of x64 servers are competitive, and they support Solaris just like Sparc boxes do, but obviously applications compiled for Sparc can't run without recompiling on x64 iron. This is not a position of strength, but it could be worse. There could be no x64 variant of Solaris, a position that Hewlett-Packard is in with its HP-UX for Itanium and IBM is in with its AIX for Power. This oversight could come back to haunt these vendors a few years from now.

Oracle was still saying this week that it can generate $1.5bn in operating earnings out of Sun in the first year following the close of the deal, and can hit in excess of $2bn in operating earnings in the second year. By itself, Sun has only demonstrated breakeven, losses, and tiny profits since its dot-com heyday. So what is Oracle to do? Probably what it always does in the wake of acquisitions, but with a hardware twist. Jack up prices on some Sun products, cut lots of overhead, kill off some projects and products, and be aggressive in other areas.

It seems fairly likely that Rock is indeed dead, particularly if it doesn't yet work. If the 16-core Niagara-3 chips work, and can be shoehorned into the Supernova chassis, Oracle might be tempted to fob these boxes off as Supernova machines.

Entry and midrange Niagara boxes will probably live on just fine, and there's no reason to believe that Oracle will stop selling Sparc64-VII machines or even do a deal with Fujitsu to peddle future boxes based on the eight-core "Venus" Sparc64-VII due in late 2010 or early 2011. It is almost certain that Oracle will try to charge a hefty premium for these boxes and at the same time position the Galaxy Solaris boxes as the price/performance leaders in the x64 market.

This will be particularly true when eight-core "Nehalem EX" Xeon processors from Intel and twelve-core "Magny-Cours" Opteron processors from Advanced Micro Devices are in the market early next year. Machines ganging up 8, 16, or 32 sockets of these x64 chips might be less elegant than a Rock design might have been, had it ever seen the light of day, but such x64 machines will have more than enough oomph to run big Solaris workloads. And there is every reason to believe that Oracle will be much more aggressive with x64 iron, in terms of scalability and pricing, than Sun has been.

Storage is anybody's guess. Oracle has said that it is keen on creating integrated systems and keeping alive disk and tape storage product lines. But it is much less clear what Oracle will change - or not change - in Sun's storage lineup. ®