Big Blue Profits, Poised For The Power9
October 23, 2017 Timothy Prickett Morgan
Let’s just cut straight to the good news. Even though IBM’s revenues have declined for the 22nd quarter in a row, and its profits are declining even faster, the substantial investments that Big Blue has made in its System z14 mainframes and Power9 systems is about to start paying off.
The System z14 mainframes made their debut in July and start shipping in the middle of September, converting from a drain on the company to a fairly large flow of cash. Power9 chip development ceased a while ago, and we presume most of the system engineering is done, and even with a slight decline in sales in the third quarter ended in September, IBM was able to significantly boost the profits it garners from its systems business. Storage was up 4 percent, which helped the top and bottom line, and offset some declines in operating systems sales, which always ebb and flow with new system sales and whatever deals IBM is making on System z mainframes to help cushion the blow of moving to a new generation of iron.
In a conference call with Wall Street analysts last week, Martin Schroeter, chief financial officer at IBM, said that System z mainframe sales skyrocketed by 62 percent compared to the year ago period, which is a pretty good jump considering that the z14 machines were only shipping for two weeks out of the thirteen weeks in the quarter. This is especially good considering that the revenue growth rate was nearly twice the sold MIPS capacity growth rate, which was up 33 percent compared to the third quarter of last year. This stands to reason since IBM was selling new z14 systems, in probably a few cases, and upgrades to existing z13 systems, in many cases. For the past several years, revenues were dominated by customers progressively activating the installed capacity latent in the iron. I have said it before and I will say it again: IBM should offer this capacity on demand capability across all machines, Power and z, not just its big boxes, and thereby provide some differentiation that X86 server makers don’t have.
That probably will not happen. But it is a good ask.
Schroeter said on the call that Power Systems revenue was down 8 percent from Q3 2016, which was not exactly a great quarter, mind you. IBM’s CFO added that sales of high end Power machinery running AIX grew again, as it has for a few quarters, and that Power iron running Linux as its dominant operating system were up by double digits, which certainly helped the situation, and said further than Linux on Power now represented over 20 percent of the “portfolio,” by which I presume he meant shipments but possibly revenue. It is probably shipments, and revenues will be a proportionally smaller chunk of overall Power Systems revenues because IBM i and AIX iron costs more for CPU, memory, storage, and probably even bent metal.
IBM is, as we have detailed many times, offering Linux shops machines with artificially lower prices to better compete with servers based on Intel’s AIX processors. We would prefer if IBM gave all Power Systems customers this pricing, but it is under such pressure to grow revenues and maintain whatever profits it can in this business that it cannot afford to be that generous. Them’s the breaks, people. And you are paying for it.
When the Power9-based “Summit” and “Sierra” supercomputers being built by IBM for the US Department of Energy for Oak Ridge National Laboratory and Lawrence Livermore National Laboratory, respectively, are accepted by the labs, IBM will be able to book a very large amount of revenue – the contracts were worth $325 million for the two machines, and a third with separate funding at Livermore, a smaller version of Sierra, will sweeten the pot. If all goes well, these machines should be operational by the International Supercomputing Conference in June 2018, and there will be a very large number put into the Power Systems column. Pieces of this deal could be accepted before then and help. But we suspect that these two machines were built at close to cost, and maybe even under it, and can be thought of as the research and development for Power9 and the related “Volta” Tesla GPU accelerators that do most of the computing in these systems. Whatever profits there are in the contracts probably come from services and software porting assistance; these machines are not generally upgraded, but replaced every three to four years.
In any event, that hasn’t happened yet, but it is a bright spot off on the horizon as far as IBM’s books go. The booking of this revenue will coincide with the revenue ramp for Power Systems machines aimed at enterprises, which most certainly have relatively high profit margins, and it will be interesting to see if the profits can grow as fast as the revenues with all of these vectors in play. My guess is not, but the amount of profit will be large and justify the substantial investment IBM has made with Power9 and will generate profits for many years to come until Power10 is on the horizon and Power9 starts to fade.
“This is strong evidence of the importance of Power to the high performance computing market,” Schroeter said. “And we will roll out commercial Power9 solutions throughout 2018. So in an era of cognitive and AI, where data is fundamental to our enterprise clients, Power is demonstrably better for Linux machine learning and deep learning workloads, the workloads of the future. And that’s why we’re committed to the Power platform.”
He did not mention IBM i and AIX, but that is because these are stable to declining businesses and the growth is all in these new workloads, all of which run on Linux. Even Microsoft, which hates Linux, has had to embrace it and use it. The Azure cloud is rapidly becoming a place where Linux runs, and even Microsoft’s virtual networking stack that underpins Azure is written atop Linux, not Windows Server.
Add it all up, and IBM had $1.3 billion in system hardware sales, up 14 percent as reported, and operating systems revenues were at $420 million, down 3 percent, for a total of $1.72 billion. IBM sold another $227 million in hardware and operating systems to other IBM divisions, and that brought the total to $1.95 billion, up 12.5 percent. Pre-tax income for Systems group was $339 million, up by 149 percent, which is good.
This is much better than IBM is doing overall. The company had $19.15 billion in sales, off four-tenths of a point, and brought $2.73 billion to the bottom line, off 4.5 percent. The next few quarters should be very good ones for Blue Blue’s Systems group, with it hitting on both the System z and Power Systems cylinders.
IBM’s real systems business, as we have pointed out many times, is quite a bit larger than the Systems group’s official numbers. There is a collection of database and transaction monitoring and middleware software that rightly should be part of Systems group and which has very little to do with Cognitive Solutions as IBM characterizes them. There is tech support, integration, and financing services that are really part of a systems deal, too, that should be counted. We do a little guesstimating, and try to reckon this, and our best shot at what that real systems business looks like is in the chart above.
As you can see, there has been a pretty dramatic downward trend here, but the curves are flattening out and, with any luck and hard work, the curves could start bending upwards soon. IBM sure is banking on that, and has done the best partnerships and engineering that was feasible to try to accomplish this. Now, we get to see how it will play out in 2018. The fate of the IBM i platform is linked to this, don’t forget.