Oracle Announces Fiscal 3Q2013 Earnings: Hardware Imploding

Oracle announced its fiscal 3Q2013 earnings. They weren't good. Even software license revenues were down.

Total hardware sales were down again: another whopping 23% (22% at constant currency) year over year to $671 million for the quarter. Hardware systems support fell less quickly to $570 million as Oracle's hardware maintenance prices continue to climb. Note that these figures include what Oracle calls its "engineered systems," i.e. Exadata, Exalogic, etc. Oracle said they showed "excellent growth" but that their average selling prices are down. Oracle did not say much about its storage business (which includes StorageTek tape products).

Oracle said that its delayed T5 processor would be introduced "next week." Oracle CEO Larry Ellison called the SPARC T5 "the fastest microprocessor in the world." He claimed that Oracle would publish 17 "world record" benchmarks. (We'll see.) Despite the claims, Oracle predicts that its next financial quarter will see another 12 to 22 percent drop in hardware sales in constant dollars (13 to 23 in reported dollars), so apparently Oracle won't be able to deliver many systems with SPARC T5 microprocessors, Oracle won't be able to sell many, or both.

Every server processor vendor has product cycles. However, if the business is healthy there are both up quarters and down quarters. We haven't seen an up quarter in Oracle's (and formerly Sun's) hardware business in a long, long time. Oracle's entire hardware revenues are now about a third what Sun's were just over five years ago, for perspective. The SPARC T4, introduced in 2011, might have slowed the hardware slide but didn't stop it. We'll see what happens next, but it's awfully hard to sustain sufficient hardware R&D with these financial results.

UPDATE: I have a few more thoughts about Oracle's "engineered systems." First, it's important to recognize that these systems are essentially "appliances": fixed function/single purpose servers that are bundles of Oracle software and rather common hardware. Exadata, for example, is an Oracle Database box. Oracle counts these engineered systems as hardware sales — they have to be counted somehow — but are they? I think they're at least as much "software" as "hardware." Read on for an explanation.

The financial problem is that every Oracle engineered system is a perfect or near-perfect substitute for an Oracle software sale. If you buy an Exadata box you don't buy Oracle Database licenses for your other servers, or at least you buy fewer of them. To some extent, probably a large extent, an engineered system sale subtracts from software sales. In other words, Oracle is simply moving many software sales among existing Oracle customers from the "software" part of its balance sheet to the "hardware" side. That could be a good strategy, or it could simply be a more expensive way to package and sell certain Oracle software products, reducing Oracle's profitability. If Oracle receives $X in profit from an Exadata sale and $Y in profit from the equivalent Oracle Database software sale, and if $X is less than $Y, then the engineered system could be a very bad idea, financially speaking. On the other hand, if the engineered systems are providing real value above traditional and cloud-based software solutions, and if Oracle can share in some of that customer value (i.e. collect more profits), or if at least the engineered systems have higher win rates with better customer loyalty, then they might be helpful to Oracle.

We don't really know yet which is the case. But there are some clues. Oracle reported that the average selling price for the engineered systems is coming down, and they say that's due to the introduction of smaller models. Even so, unit sales are below previous forecasts, even the revised ones, and in this most recent quarter software license sales are down. I wonder whether there might be some cannibalization going on, with the modest engineered systems sales understandably eating into Oracle's software sales. Margins are much higher on software, so if cannibalization is happening it's probably not a good trend. Also, it's hard for me to understand how fixed function appliances fit into the growing public and private clouds. Yes, I'm sure Oracle will have some marketing answer why their engineered systems are compatible with the word "cloud." I don't agree even if I stretch the definition of "cloud," and I suspect a lot of potential customers would also disagree with Oracle, viewing their engineered systems as what they probably are at least in part: vendor lock-in devices. If you only want to buy your storage from Oracle, for example, buy an Exadata box. (And databases never get larger, of course.)

I haven't seen IBM follow this path. And to the extent Oracle thinks it's following in IBM's path, Oracle is misreading IBM and its history. Yes, one can buy WebSphere Application Server appliances for example, but they are virtual ones with plenty of convenience and also lots of flexibility. IBM's "hardware" appliances tend to be very specialized with no direct software-only counterparts. The IBM PureData System for Analytics and DataPower are two excellent examples. My observation is that, in general, IBM is trying to deliver "workload-optimized" systems which are extremely convenient, easy to deploy, and easy to manage, but which also have a lot of flexibility. Granted, IBM has a lot of advantages, including the fact that IBM can research, design, and fabricate its own silicon. I can't think of any other big enterprise software vendor that can do that. As I've written before, it takes a considerable amount of long-term effort with sustained investments to make that hardware-software magic happen, and it's something IBM does uniquely well in its markets.

by Timothy Sipples March 21, 2013 in Financial
Permalink | Comments (0) | TrackBack (0)

The postings on this site are our own and don’t necessarily represent the positions, strategies or opinions of our employers.
© Copyright 2005 the respective authors of the Mainframe Weblog.