Reality Reflects IBM i, Which Reflects It Back
January 20, 2020 Timothy Prickett Morgan
In the fullness of time, the datacenters of IBM i shops will reflect the trends in the overall IT sector and in the respective industries and IT scale of their peers who have other back-end systems of record and a mix of systems of engagement. Give or take.
That’s why we pay attention to the larger trends going on in the IT sector. The IBM i base may be coming a little slowly to true cloud compute, storage, and networking capacity, but it is getting there finally thanks to the efforts of IBM, Connectria, Skytap/Microsoft, Google, and others.
It takes a while to sort through all the numbers for infrastructure sales and figure out how much was for public and private clouds and how much of it was for traditional, bare metal infrastructure. And we are not sure how IDC qualifies machines with just logical partitions or virtual machines but not cloud management and utility pricing or chargeback, so we think there is a big middle ground that could not be properly counted. But IDC has come up with its best estimation of how spending on servers, switching, and storage combined split across public clouds, private clouds, and traditional on-premises gear in the third quarter. And for the second time in history, aggregate spending on cloudy infrastructure was slightly larger than for non-cloud, traditional servers, switching, and storage.
Spending on infrastructure by the myriad public cloud providers fell by 3.7 percent to $11.9 billion in the third quarter of 2019, but sales on the same gear going to make up private clouds at enterprises in their own datacenters or colocation facilities rose by 3.1 percent to $5 billion. Add it all up, and cloudy infrastructure accounted for $16.8 billion, down 1.8 percent. That may not sound like good news, but the growth sequentially is quite significant and it shows that spending is on the rise after a bit of doldrums earlier in 2019. Take a gander:
That said, as you can see in the chart above, spending on non-cloud IT gear is still trending downward, but at no worse of a rate than happened throughout 2014 through 2016 and at slightly better than the spending levels seen throughout most of that time. As you can see from the chart above. Non-cloud IT spending has been generally rising, with some pretty intense sawtoothing, from 2016 though 2018 and took it on the chin a bit in the first three quarters of 2019. This non-cloud IT infrastructure spending was down 7.7 in Q3 2019, to $14.7 billion. This part of the business seems to be on its own cycle and it will be interesting to see if it rebounds in Q4 2019 and holds up in 2020. By 2023, IDC is projecting that this non-cloud IT infrastructure spending will only account for 41.9 percent of the infrastructure pie, down from 51.6 percent in 2018. So the prognosis is not a particularly good one. It’s not bad, either.
But the slow nature of this transformation from bare metal or relatively lightly virtualized systems to true cloud infrastructure, whether public or private, will take two or three decades. And it may never full convert, and in fact, it may reverse. There are plenty of systems that work on bare metal and give a lot of the flexibility of a stack of machinery with virtualized server, storage, and switching without having to add all of that overhead on performance or cost. I have no idea where IDC is counting sales of such bare metal clusters, but for certain kinds of data analytics, scientific and financial simulation and modeling, and artificial intelligence applications, bare metal is the way to go and there is no way companies are going to add virtualization. That said, IDC says that 2020 is the tipping point where non-cloud IT infrastructure revenues worldwide go below 50 percent and stay there for good.
For all of 2019, IDC is projecting that servers for cloud-style IT will account for $34.1 billion in sales, down 3.1 percent. Spending on storage underpinning clouds will rise by a mere eight-tenths of a point (we don’t know how much money, IDC doesn’t say and there is no way to calculate it from its public statements), while Ethernet switching for clouds will rise by 11.2 percent.
My best guess is that these strategies are not an either-or proposition, and that companies will continue to do a mix of all three types of IT infrastructure consumption. In the case of IBM i, it might be test/dev and disaster recovery in the cloud, mission critical database and ERP on premises, and a mix of adjunct artificial intelligence and data analytics services for systems of engagement that run as SaaS products but have cloudy IT infrastructure or automated bare metal instances underpinning them. Probably in the proportions we are seeing emerge above, in fact.