After Seven Quarters Of Growth, Power Systems Declines
October 21, 2019 Timothy Prickett Morgan
The tough compares have hit home on IBM’s Power Systems business, but the good news is that this has happened after seven consecutive quarters of growth for the Power-based server business that Big Blue owns lock, stock, and barrel. Even with this decline, which was quite steep because of the triple whammy of tough compares (more on that in a moment), there is still a healthy underlying Power Systems business that is much better off than the last time it was hit by similar declines.
Let’s take a look at the numbers for IBM’s Power Systems division and then work our way up through its Systems group and to the company at large. According to the presentation put together by IBM’s chief financial officer, Jim Cavanaugh, to go over the numbers for the third quarter of 2019, the Power Systems division had a decline of 27 percent in constant currency (meaning growth in local currencies aggregated across those economies), with as-reported sales also being down 27 percent. In other words, currency had no effect on the overall Power Systems business even if it did impact IBM’s sales, as reported in U.S. dollars, by 1.3 percent in the period ended in September.
If you plot out the growth and decline rates for the Power Systems business since the Great Recession ended in early 2010, here are how the rises and falls look:
That’s a pretty unnerving chart, I realize, with a lot more time spent below the neutral line than above it. But the good news, if you plot it out with the launches of the Power7 processors in 2010, the Power7+ processors in 2012, the Power8 processors in 2014, and the Power9 processors in 2017-2018, the growth rates have been steadiest and largest during the Power9 generation. Power7 did better than Power8, and Power7+ didn’t really do much at all to spur growth, which is why you didn’t see IBM launching Power8+ or Power9+ processors between its cycles and why we suspect there won’t be a Power10+ or Power11+ kicker, either. (I could argue why IBM should do this. But not today.)
Earlier this year, we built what we think is a pretty good quarterly revenue model for the Power Systems business at IBM, and here is what these growth and decline numbers translate into as quarterly system sales for machines based on Power chips:
As we have pointed out before, the Power Systems business used to be considerably larger, and has been shrinking steadily from its highs in the late 1990s. The dot-com bust was not kind to the RISC/Unix business in general, and the AS/400 market in 1999 consumed a vast amount of capacity for Y2K projects that took many years to burn off. The relentless pace of Moore’s Law kept increasing processor performance (as measured in CPW throughput) for the past two decades faster than businesses were growing their revenues and their transaction volumes. Part of the decline in the Power Systems business is just this mismatch between the growth companies have in their core business applications and the performance increases IBM delivers to chase other workloads, such as in the HPC and AI markets today. This is good for customers, but it is not always good for IBM when the typical IBM i shop has one or two, maybe sometimes three or four, cores running IBM i and that’s it.
The good news is that, despite all of the pressures, the Power Systems business has been growing for the past seven quarters, and while there is always a chance that IBM will turn in a killer fourth quarter to save the day – or rather, year – for 2019, we think this is probably highly unlikely. If IBM does about the same revenues in the fourth quarter as it did in the third quarter, then it will turn in about $1.6 billion in core Power Systems revenues compared to just a tad over $2 billion in 2018, according to our model. We think Q4 will probably see the normal seasonal uptick, of around 50 percent compared to Q3 in the past two years, and that would mean this core Power Systems business could hit $1.75 billion in sales. That would still be a 13 percent decline for the full year, but considering that IBM booked $325 million in 2018 for the Summit and Sierra supercomputer deals, if you take that out of the mix, then the underlying business will actually grow by 3.5 percent. That does not include something on the order of $650 million to $700 million in revenues that comes from systems software sales on Power iron and upgrades for systems that Global Technology Services installs on behalf of its outsourcing customers.
These sales figures also do not include the cost of custom Power9 gear built by Google and Rackspace Hosting for internal use, or the Power iron sold by Inspur and Supermicro. Some of the reason that IBM’s Power Systems revenues have declined in recent years is because it has essentially been locked out of the Chinese market, which was a fast-growing market for IBM i. The Chinese government wants to have indigenous suppliers for mission critical systems, and Inspur alone probably now accounts for something on the order of $500 million in annual Power iron sales all by itself. In years gone by, that would have been money that Big Blue booked. But now it can’t.
IBM could really have used one or both of those exascale deals from the U.S. Department of Energy, but the core business is still growing, however minutely in 2019, and there is a chance that IBM can and will do something to push the Power Systems platform more aggressively in the coming year to get more money into its coffers for its indigenous servers.
The irony is this: The overall core Power systems – lower case S there – business is actually probably a little healthier in 2019 than it was in 2018, and is for sure healthier than it was in 2016 or 2017 and is on par with the $2.3 billion in revenues for core systems (meaning just the server) revenues in 2015.
As for IBM, the “real” systems business that it has, including System z mainframes and Power iron plus their systems software, services, financing, and other stuff, is essentially flat year on year, as best as we can figure, and the decline is nowhere as bad a many think by looking at the foolish way IBM presents itself to Wall Street. And estimate by us includes the $371 million in revenues from Red Hat that was recognized by Big Blue in the quarter and that went into the Cognitive and Cloud Software division. It is hard to draw the lines between Linux and the OpenStack cloud controller and the OpenShift container controller (which is based on Kubernetes), so we get why IBM is not putting Red Hat revenues in the Systems group. But we have allocated this back to the “real” systems business in our model, as best as we can.
Let’s go through the numbers here. IBM sold $1.1 billion in systems hardware in Q3, down 17.2 percent because of the double whammy declines in both Power Systems sales (off 27 percent) and System z mainframe sales (down 21 percent). Based on IBM’s presentations and our math, we reckon that IBM brought in another $380 million in Q3 for operating systems on its own iron, down 8 percent. Storage sales were off 6 percent. IBM’s total Systems group revenues – including $195 million in stuff sold to other IBM groups – was down 15 percent to $1.67 billion. This Systems group business had gross profits of $785 million, and pre-tax income of a mere $39 million because of the substantial investments IBM is making in future technologies for systems.
Overall, IBM’s revenues were down 3.9 percent to just a bit over $18 billion, and net income – somewhat more alarmingly – fell by 37.9 percent to $1.67 billion.