The Power Systems Decline Did Not Have To Be This Bad

April 24, 2017 Timothy Prickett Morgan

Thirteen weeks ago, when IBM reported its financial results for its final quarter of 2016, we said that the Power Systems business would decline in every quarter ahead of the Power9 launch, and while this is not a hard thing to predict, it has certainly turned out to be true in the first quarter of 2017. Both IBM’s Power Systems and Systems z machines and their related operating systems were hit by gut-wrenching declines.

The good news, as we have pointed out time and again, is that IBM’s true systems business is distinct from and is a superset that is much larger than the underlying server business that it reports for its Systems group. But the situation is a little more subtle than even this.

On the one hand, the IBM i platform represents the largest customer base that that company has, and certainly as large as the AIX and Linux on Power and System z mainframe bases put together. We have not seen data on the installed bases of Big Blue in so long it is hard to project from where we were in the early 2000s until now. I think IBM has around 120,000 customers that are running IBM i, i5/OS, and OS/400 systems in the world today, and maybe 30,000 of them are very active customers who keep relatively current. IBM has somewhere on the order of 5,000 to 6,000 mainframe customers, and by definition, these are active customers because they have to pay monthly licenses for their software stacks even if their hardware is ancient. IBM used to have around 100,000 RS/6000 customers running AIX systems (not workstations), but I think many of these shops have migrated to Linux for infrastructure and application server parts of their workloads and that AIX remains (as does Solaris and HP-UX) largely as a database platform at the heart of transaction processing systems.

The shape of the bases stands in stark contrast to IBM’s businesses. As far as we know, IBM is not making a concerted effort to get IBM i shops doing machine learning, sophisticated distributed data analytics, high performance simulation and modeling (useful for manufacturers, of which the IBM i base has plenty) and their analogs in the financial services and insurance sectors. Way back in the day, when I was cutting my teeth as a journalist and analyst, when IBM had a hot new technology, it made sure it had play on all four of its platforms – AS/400, RS/6000, System/390, and Netfinity as they were called a long time ago. IBM pushed Smalltalk object-oriented programming out to all of its platforms long before Java came along – remember that? – and time and again was pushing advanced technologies to customers.

The genius of the AS/400 platform is that it had many of its own clever technologies, unique to itself like an easy-to-use relational database, as well as those technologies that spanned the four platforms. But it was a member of the family, and more importantly, IBM had a vision for all of its customers, large and small, and then it executed on that vision as appropriate for those distinct customer types and platforms.

Here’s a test: Tell me what IBM’s vision is, and what you say has to apply to companies other than the Global 2000.

Yeah, that’s what I thought. IBM is so busy trying to transform itself that it forgot that it had a responsibility to transform its customers, even small and medium businesses – companies that bear very little resemblance in terms of budgets and technology sophistication but who want to be modern and transformed just the same. And, by the way, I am not talking about changing from green screens to GUIs for tablets and smartphones here. Other companies, many of whom support our publications, as well as dozens of key individuals on the IBM i community are doing their best to bring new technologies and techniques to the platform, and we commend them for that. But it is Big Blue’s job to do that, and to make it economically possible as well as technically possible for them to consume these new technologies and grow and change their businesses.

This is what Amazon Web Services does. This is what Microsoft does with the Windows Server stack and its Azure cloud extension.

If IBM wants to revitalize its platform business, it has to do more than port Linux, Hadoop, Blockchain, and whatnot to each of its platforms. This is a first step, but it is only necessary, not sufficient. And given this, it is no surprise whatsoever that the Power Systems business and the mainframe business are both utterly dominated by their own product cycles in a way that is not particularly healthy. Intel doesn’t have this problem with its Xeon chip business because demand keeps growing for core compute because nearly an entire IT ecosystem runs on its chips these days. Intel has worked very hard to get to this point over 25 years, and it is fighting like hell to maintain it. IBM could take a lesson here, and if it did, starting with the Power9 launch, it could foment a resurgence of the Power Systems platform alongside the hard work that it is doing in the OpenPower Foundation. We remain hopeful that hyperscalers and HPC shops will lead the way and help get Power growing again.

What is clear is that IBM cannot do it alone, any more than Intel could in the early days. (The Xeon tail is wagging the server dogs in the datacenters these days, however.) In the first quarter ended in March, IBM sold a little more than $1 billion in system hardware, by which IBM means Power Systems and System z servers as well as disk, flash, and tape storage and a smidgen of networking that it resells, and as best as we can figure, this part of the business was down 18.7 percent but still had a gross profit of $322 million. Operating systems raked in another $385 million, down 10.9 percent, and gross profits for this part of the business came to $335 million. The total revenue for the Systems group was $1.56 billion, and total gross profits came to $657 million, which sounds good, but after all of the overhead is allocated excepting taxes, Systems group posted a pre-tax loss of $186 million.

The System z hardware business was down 40 percent in the first quarter of 2017, and the Power Systems business was down 27 percent – and that was despite a growing Linux-on-Power business. IBM’s flash arrays took off and helped lift its overall storage business by 7 percent in the quarter. Which is good news after a long wait, but again after years and years of decline in storage that makes for an easy compare. The compares for late 2017 and early 2018 after the Power9 systems start shipping in volume will also be easy because of these big drops. Admittedly, both the System z13 mainframes and Power8 midrange and high-end boxes are at the end of a very long tail that is now three years old, so that is part of the problem. And as we have pointed out before, the real IBM systems business – including all of that middleware, tech support, financing, and such – is much larger and did not take such a big hit.

As far as we can calculate, this true base systems business accounted for $5.44 billion in revenues in the first quarter, down 7.6 percent year on year. As you can see from the chart above, there is a correlation between this systems business and IBM’s overall sales. But the underlying systems business is falling faster than the “strategic imperatives” that IBM focuses on, including Watson machine learning and other cognitive technologies that the company has put an enormous amount of money and effort into. It seems to me that the real strategic initiative that IBM keeps missing is that its one and only strategic initiative is to make these technologies relevant to the existing customers it has. This seems obvious to us.

One thought on “The Power Systems Decline Did Not Have To Be This Bad”

Greg says:

November 30, 2017 at 6:23 pm

I like the info here but the AS400 didn’t really have an easy to use relational DBMS because developers didn’t define relationships between their tables. It was a non-relational DB for all intents & purposes.

Of the 30,000 active and current customers, how many would you say are writing and maintaining in-house apps in RPG? How many in Java, PHP and Node?

This Issue Sponsored By

Table of Contents

Content archive

Recent Posts

Subscribe

Pages

Search