Analysis It is hard to believe that three whole years have passed since software giant Oracle instantly became a systems player by snapping up beleaguered Sun Microsystems for a cool $7.4bn – about $5.6bn net of Sun's cash hoard.
No one took El Reg up on our suggestion to start up a Sun clone we called Moon Macrosystems by grabbing all of the open source goodies (including chip designs) to create integrated Sparc and X86 systems running Solaris and the rest of Sun's considerable software stack, denying Oracle the control of Java and Solaris that it sought.
That's a damned shame, because it would have been a lot more fun.
But let's face it, if Sun had been acquired by IBM, which was the original plan back in November 2008, it would have been a lot less fun.
Oracle cofounder and CEO Larry Ellison, as it turns out, has caught a bit of Big Blue hardware bug and actually seems to like the idea of peddling "engineered systems" to enterprise customers who know Oracle mostly for its database, middleware, and application software.
So, was Oracle's Sun acquisition worth it, and has it created something that will live and grow? Ellison certainly wants everyone to believe that, and has told Wall Street and his investors that he believes that. And if you look at the Sun deal from a lot of different angles, it's hard to show that it was not a good deal for Oracle – or for the egos of its top brass.
To understand why, it helps to remember the context in which Sun Microsystems' former CEO, Jonathan Schwartz, started shopping the formerly upstart Unix workstation and server maker and Java creator to IBM.
In late 2008, the Great Recession was just starting to build momentum – although we may not have known how truly ugly it was going to get – and during recessions, spending on systems is usually one of the first things to take a dive. Usually, sales of X86 iron get frozen first, when possible, but big Unix and mainframe systems, which endure a long process during an upgrade, continue with their budgetary inertia and deals proceed.
This is great at first, but eventually you'll want to do new deals for Unix and mainframe servers - and customers will be slow to rebuild that momentum.
Moreover, recessions tend to cause system transitions. It's no coincidence, we can now surmise, that IBM was talking to Sun just when Intel was prepping its "Nehalem" Xeon 5400 server processors and Cisco was about to launch a new category of modular systems – and itself – into the server business that following March.
While HP and Sun were both selling blade servers with some integrated and virtualized networking and I/O, Cisco came up with a new form factor and bootstrapped itself into the server racket from a standstill. Now, nearly four years later, it's jockeying for position to get into the top five – if it can grow enough to knock out Japanese server giant Fujitsu, that is. It may take a while longer, but Cisco could do it.
Sun's own "Constellation" integrated systems were well regarded and had a fair amount of engineering compared to standard rack or blade servers. But Sun's initial and exciting foray into x86 servers, driven by its early and enthusiastic support for AMD's Opteron processors a decade ago and widened by the adoption of Xeon processors from Intel, eventually ran out of gas.
Sun could not increase x86 server revenues fast enough to make up for the ground it was losing to IBM in the Unix system biz, even if its overall server shipments increased radically. Sun tried to make it up in volume – by selling more stuff and by making more noise by open-sourcing everything, er, under the sun.
That strategy really didn't work all that well, although it made for lots and lots of interesting times in Server Land.
Not surprisingly, Oracle's acquisition of Sun has made for a continuation of interesting times in the Land of Tin and Iron. And thank heavens for that, because without Oracle being a gadfly as Sun always was, the rest of the nobles in Server Land might be boring us all to death. Say what you will about Larry Ellison – seriously, that's what Comments are for – but he is almost always interesting, he is often right (even if he rewrites history a bit here and there), and he has the money to prove it.
Which brings us to the first point to consider about the Oracle acquisition of Sun three years ago? Did this deal pan out? Did it make financial sense?
On Monday, El Reg asked Oracle co-president Mark Hurd – the former HP chairman and CEO who had a chance to buy the Sun server and storage businesses when all Oracle wanted was Solaris, Java, and MySQL, but turned it down – those very questions. And Hurd was unequivocal about the Sun deal having paid for itself.
"Yes, materially so," Hurd explained. "If you look at something like cash flow from the deal in relationship to the purchase price, the cash flow has far exceeded the purchase price of the acquisition."
We have to take his word for that because Oracle does not report cash flow from any of its divisions and segments – more on that in a minute.
The point of the acquisition of Sun was to instantly make Oracle a systems player – and to give Oracle some leverage in the data center where it was previously dependent on other's hardware to support its products. And to that end, Mr. Hurd, what has the effect of the Sun deal been on Oracle?
"Phenomenal, because it had multiple dimensions," he said. "There's a Java element, which has a financial element as well as the enablement piece of Java as it relates to our development activities. Point two, obviously the support base that it gave us. Point three, the hardware revenue stream that was extremely positive and accretive in addition to the platform on which it gave us to develop Exadata, Exalogic, etcetera.
"So [it was a success] financially when you look at the acquisition price compared to the cash flow that has been generated, and second the development tools and foundation it has given us to build a whole new series of technologies."
It is not easy to do a before and after comparison of the Sun Microsystems business and the bits of it that shine through the Oracle financials. That's not just because Oracle is obfuscating, but because bits of Sun are spread all over the place inside of Oracle these days.
Oracle sells Sun systems and books their revenues in different ways, as well, which complicates things further. Sun broke out computer systems product sales and Solaris sales separately as it was getting eaten by Oracle, and put Solaris support revenues in its services segment. Oracle bundles the Solaris and Linux operating systems with its machines (not just a physical bundling, but also in the price) and sells hardware support contracts separately.
Let's take a look under the hood
For the purposes of a thought experiment and to see if what Hurd said about the Sun deal paying for itself is reasonable, El Reg took the last couple of years of financials for Sun and plotted them against the hardware systems and support businesses of Oracle.
Sun's overall revenues came mostly from systems, and most of Oracle's Sun product sales are also mostly for systems, but obviously they are not perfect apples-to-apples comparisons. But the data is interesting nonetheless.
Here's what the revenue curves look like for the two companies, with the quarter of sales at the end of 2009 – before the acquisition was done and before Sun has reported its second quarter of fiscal 2010 financial results – blocked out. The comparison also ignores the first month of Sun product sales at Oracle, which closes its third quarter in February each year. Check it out:
The revenues of Sun in total, and Oracle for hardware systems and support for them
Sun was declining on its own – and perhaps intentionally for all we know – starting in November 2008 when it first approached IBM to try to stem its losses from the low-margin x86 server business. You can see that Oracle continued Sun on this trend line, doing absolutely intentionally what Sun itself might have best done by itself had it realized that it could not be all things to all data centers.
Oracle promised to sell Sun's own products backed by its own intellectual property and to jettison unprofitable storage products to get out of the volume-server biz where there are no margins unless you're Intel, Microsoft, or Red Hat.
Oracle has chopped back those bad businesses even as it has built up new "engineered systems" businesses based on the Exadata database cluster, the Exalogic application server cluster, the Exalytics in-memory database machine, and the Sparc SuperCluster – a kind of database-application server hybrid based on Sparc T4 server nodes and the flash-pumped Exadata storage servers that give the database cluster its name.
As you can see, Oracle figured out the hardware systems support business right away. It is rock steady, and probably was fairly steady when Sun was running it, too.
Hardware revenues have been declining, along with sales for all of those engineered systems as well as for free-standing Sparc T4 systems that saw double-digit growth in Oracle's most recent quarter. Hurd said that ZFS storage array sales are also growing, and that there was over 70 per cent sequential growth in the quarter for engineered systems, with over 700 systems sold in the quarter. The engineered-systems business tripled (that's shipments, not revenues) in Oracle's prior fiscal year and was on track to double this fiscal year. As Oracle's top brass said back in December, the company is committed to growing this hardware business again.
"We're right about at those crossroads," said Hurd. "We don't pinpoint it to within a week or a month, but we're right in that wheelhouse where right in the next period of time you will start to see complete growth for the top end of the business overall."
We think Hurd meant to say "top line," not "top end." It doesn't make sense otherwise.
It gets interesting when you look at the gross income for Sun, on the left, and Oracle's hardware systems business, on the right:
Oracle has been able to smooth out some of the choppiness in the gross income of Sun systems biz
What you can't see in those gross income numbers above – which take out the cost of providing the products sold by Sun and Oracle but do not include sales, marketing, research, development, and other administrative costs as well as pesky things like writeoffs and taxes – is that Sun took a huge wonking write-off while the Oracle deal was pending, shedding nearly 6,000 workers and writing off the value of some acquisitions. This was done to gussy up Sun's books before the company was to be absorbed by Oracle, should the governments of the world approve the acquisition.
But you can see that write-off and the fact that Sun was never able to make much money in this chart, which shows the operating income for Sun as a whole and the Oracle hardware systems biz:
Operating income at Sun versus estimated operating income for Oracle's hardware biz
Oracle reports its operating income only as a whole, not for individual product lines, and the chart above reckons that the operating income for hardware eventually settles out to be roughly proportional to Oracle as a whole. This is what Oracle's stated goal was, more or less, when it acquired Sun.
If you do some estimating for sales and gross and operating incomes in December and January at Oracle for the hardware biz, and then do some math, here is what the numbers look like: since taking over Sun in the last week of January 2010, Oracle has done $11.72bn in hardware system sales and $7.37bn in hardware-support sales, for a grand total of $19.1bn. Gross income from the Sun hardware business, after it has been pared down by Oracle, was just under $10bn, and if our operating-income estimates hold, they come to just a smidgen below $6.7bn. So at the moment, Oracle has gotten its $5.6bn back plus another $1.1bn.
That's not too shabby.
Moreover, if Oracle had decided to get into hardware and had to bootstrap itself into servers, it would have had to spend countless billions on research and development, learning how to do a hardware supply chain, and then marketing the hell out of the boxes. In actuality, all Oracle had to do was say, "Larry Ellison is a grown up and he is now going to fix the Sun business like Scott McNealy either couldn't or wouldn't."
Sun's problem was that it was more of an offshoot of Stanford University, and that like IBM in the late 1980s and early 1990s had an expensive research organization that was paid for by a monopoly that was drying up. For IBM it was the mainframe and at Sun it was the Sparc/Solaris platform that was, for about a decade, one of its heirs and one that made lots and lots of money for Sun.
But when IBM came fighting back with Power Systems, boosting performance and getting very aggressive on price – which IBM could do because it had a vast base of proprietary systems that are stickier than the La Brea tar pits – Sun acted like nothing had changed. Just like IBM did for about the same amount of time when the mainframe business started collapsing thanks to Unix servers.
Oracle is a vastly more efficient operation than Sun was, and it doesn't wander off and spend hundreds of millions of dollars a year on science projects that may or may not pan out. The company spends plenty on R&D, but it is not trying to match the patent portfolios of IBM. Oracle sees an adjacent market and eats enough players to become either the largest or second-largest player.
That is probably not going to happen in systems – but you never know.
After all, the stickiest thing in the world is a database housing all your vital business data. As long as customers love Oracle's database, and Oracle makes the systems that run it best, Oracle has every bit as good of a chance of ruling the enterprise data center when it comes to systems as does upstart Cisco Systems, which will be half Oracle's size in systems if 2013 plays out as expected.
The difference is, Cisco is still growing like crazy and Oracle is still shrinking. Cisco could level out with a systems business somewhere around $3bn to $4bn a year several years hence. Oracle is already there (and I am just talking about hardware, not support), and could get back to maybe $5bn.
Oracle has staked its game on integrated systems, in owning the entire customer experience from disk drive to application screen, and thus far it is the only vendor who can claim this. The IBM of the 1970s and 1980s, which used to develop its own application software as well as databases, transaction monitors, and so on, is the last time any company had this much control over a stack.
The question is, will customers want to be locked so tightly into Larry's World? ®