Itanium Dump Could Cost Oracle Billions

Mar 25, 2011 (11:03 AM EDT)

Read the Original Article at

Larry Ellison thinks he has found a way to shore up the struggling server business he bought for $7.4 billion last year in the form of Sun Microsystems—force customers to buy Sun hardware if they want to continue to run the latest Oracle software. That's about as close to a bet-the-company strategy as you can get.

A quick recap: Oracle said last Wednesday it will discontinue developing versions of its software for systems that run Intel's 64-bit Itanium chip. About 50% of HP Itanium-based Superdome servers in the field are currently running Oracle, Forrester estimates.

Oracle said its decision stems from its belief that Intel isn’t committed long term to Itanium, which the chipmaker co-developed with HP and finally introduced in 2001 after a series of production delays. Oracle's very public questioning of Intel’s and HP's honesty around Itanium set off one of the nastiest flame wars Silicon Valley has seen in recent years.

Oracle accused Intel and HP of "knowingly withholding" key information about their Itanium plans—i.e., that they're on the verge of ditching the platform in favor of a pure x86 strategy. In turn, a top HP official accused Oracle of engaging in "anti-customer behavior." Intel CEO Paul Otellini said he felt compelled to reaffirm Intel's commitment to Itanium "as a result of recent announcements from Oracle."

These hand-fed quotes are great fodder for bloggers with daily deadlines to meet, but the heat:light ratio from all this is about 1,000 kelvins:one hungover firefly. CIOs who have committed millions of dollars to servers from these vendors expect more than a bunch of name calling that proves only that the guys running these tech giants really are still nerds at heart—yachts and private jets notwithstanding.

Here's what's going on behind the PR wall and, more important, what to expect. Oracle isn't moving as much Sun hardware as it needs to in order to justify the billions of dollars it paid for those assets. That's not just my opinion. Oracle on Thursday released its fiscal third-quarter earnings, which showed that new hardware sales, which came in at about $1 billion, missed earlier guidance from the company by $100 million.

To boot, sales of Oracle/Sun servers declined 14.4% in the fourth quarter of last calendar year, according to IDC, despite the fact that the overall server market grew 15% during the period, as IBM server sales rose 22% and HP’s rose 13%.

Given all this evidence, it's pretty clear what's really behind Oracle's decision to dump Itanium. In the enterprise market, the database-application suite pairing drives the hardware decision, and since Oracle is still far and away the market leader in database-plus-ERP software, it has decided to try to leverage that position to boost its flagging Sun business.

In other words, moving forward, customers that want to run enterprise-class installations of Oracle software will have to do so on Oracle's enterprise-class hardware. I'm not suggesting that Oracle will end support for smaller, one-off installations of its software on x86 platforms, but for the top line, zero-latency stuff for banks, big manufacturers, research institutions, and other demanding environments, it's going to be Oracle-on-Oracle.

Oracle hasn't made any specific announcements to this effect (other than its public diss of Itanium), but the writing is on the wall. What other high-end platforms are there?

You can run Oracle on an IBM AIX/Power7 architecture, but given the open hostility between Ellison and Big Blue and their intense competition in databases, it wouldn't be surprising if AIX is the next platform Oracle will stop supporting. There's still Dell, of course, but Dell's Oracle platform, based on Intel Xeon, doesn't go beyond the x86 world. That's not enough for many big-time customers.

The bottom line is that Oracle is heading toward becoming a propriety systems vendor, confident that CIOs will cast aside their aversion to single-suppler lock-in in order to continue running Oracle’s most advanced software stack.

It's a risky bet—one that could end up costing Oracle billions in future revenues from license sales and support. Forcing current Oracle-on-HP Itanium customers to spend what could be millions of dollars to port their entrenched business apps to the Solaris/Sparc architecture won't sit well with most of them.

CIOs of multibillion-dollar companies aren't used to being told what to do by vendors. And given the time, energy, and hassle involved, many would view the transition as an opportunity to consider a brand new setup, like, say, replacing Oracle PeopleSoft ERP with SAP paired with IBM's DB2 database and AIX operating system on Power7 hardware. They could also turn to the cloud, and leave the integration headaches to SaaS vendors like Salesforce.

On the other hand, Ellison doesn’t have much choice but to take his best shot. He loaded up on $7.4 billion worth of Sun chip fabs, assembly lines, research facilities, design centers, mouse pads, and coffee cups—now he's got to move that big iron, somehow.