The Cloud Capex Conundrum

MTailor: Cloud-Native Jeans
November 15, 2019
Cloud Computing and Application Inertia
December 19, 2019

A few days ago I came across this very interesting chart on Twitter:

Don't miss a single blog!

Don't depend on Twitter, Facebook, or Linkedin notifications. Sign up and get the latest blogs and videos delivered to your inbox asap!

I agree to have my personal information transfered to MailChimp ( more information )

I will never give away, trade or sell your email address. You can unsubscribe at any time.

The chart comes from IDC’s recent “Worldwide Semiannual Public Cloud Services Tracker, ” and depicts public cloud OPEX (i.e., cloud provider revenues) and “Traditional IT and Private Cloud” enterprise capital investment for the period 2016 – 2022. In other words, how much are users spending on cloud services versus how much are they investing in their own computing infrastructure?

IDC is very well known for its quantitative research based on extensive surveys of users and vendors. For this Tracker report, it covered companies working in general cloud computing (IPSaaS) as well as verticals (e.g., AI), located in geographies across the world. Overall, IDC addressed 53 geographies and 80 discrete market segments.

There are a couple of things that jump out from the chart.

Public cloud use is exploding

IDC predicts that over the seven years from 2016 to 2022, public cloud spending will jump by more than 500%. By 2022, users will be spending more than $100 billion on public cloud. This is consistent with AMG financial results I discussed in a recent post. This is clearly the fastest adoption of a new technology in the history of the industry. It’s easy for this critical fact to get lost in the day-to-day noise of product announcements, contract disputes, and adoption challenges. But let me repeat that statement: This is clearly the fastest adoption of a new technology in the history of the industry. And from the slope of IDC’s chart, it doesn’t look like public cloud adoption will slow down in out years beyond 2022.

Supporting this, of course, is massive capital investment, which I’ve written about in the past. Cloud providers the world over, and especially AMG, are spending tens of billions of dollars annually on building facilities to support the massive user adoption IDC’s chart describes.

It’s important to understand the impetus behind public cloud provider revenues and capital investment: application deployment. Users are deploying tens of millions of applications into cloud infrastructure, drawn by cloud computing characteristics unavailable in the long durations and upfront investment associated with traditional infrastructure.

This represents a massive movement within the industry. Public cloud computing is clearly a critical component in the software value chain. Moreover, what we think of as an application is itself morphing, evolving in response to those characteristics noted above. Cloud-native applications are partitioned into microservices, respond dynamically to changing workload, implement performance measures to improve user satisfaction, address much larger amounts of data, and implement new techniques of monitoring and management.

So, one clear implication of IDC’s chart is that massive adoption of cloud is occurring, driven by ginormous numbers of applications being deployed there, and that trend will continue for the forseeable future.

Private Infrastructure is Not Going Away

Here’s an interesting thing, though: all that public cloud adoption does not appear to be coming at the expense of private infrastructure. IDC’s chart shows private infrastructure holding steady at around $75 billion. Despite the huge adoption of cloud computing, enterprises are clearly maintaining their on-prem environments and continue to operate applications there.

Frankly, this is a bit of a surprise to me. I have been involved in cloud computing for well over a decade, and always expected that there would eventually be a significant migration out of data centers and into the cloud, based on the manifest advantages of public cloud environments.

Clearly, I was wrong. 

In fact, despite high profile companies like Expedia, Dow Jones, and Time Inc., most enterprises have left much of their existing application portfolio in their data centers.

One question might be whether they will continue this practice. After all, the IDC chart extends only to 2022. Given the steady level of investment over the entire span of the chart, it’s unlikely that there will be a significant reduction in on-prem investment for the foreseeable future — a fact that carries important implications for new applications, a topic I will address in my next blog post. The clear message of IDC’s chart, though, is unmistakable: on-prem computing environments will remain present and serve as the computing infrastructure for significant portions of most enterprise legacy applications.

The explosion of application portfolios

So, the message of the IDC chart seems to be: 

  • On-prem isn’t going away and is in fact holding steady. Enterprise application portfolios are, by and large, remaining in the infrastructure environment in which they were originally deployed
  • Public cloud use is exploding, growing at 40%+ year on year. Millions of applications are being deployed there to take advantage of the agility and rich service offerings characteristic of that infrastructure environment

The obvious conclusion from these two elements is that the overall size of enterprise application portfolios is growing rapidly — according to Gartner, global IT spending is growing at around 3% per year on a base of $3.8 trillion, which works out to around $100 billion, or roughly the amount being spent on public cloud services.

Put another way, worldwide IT spend is growing significantly, and all of the incremental spend is going toward public cloud environments.

The reality of enterprise IT infrastructure in the future will be maintaining a significant on-prem environment, along with the legacy portfolio applications residing there, and deploying a rapidly growing portfolio of cloud-based applications leveraging the public providers’ infrastructure. However, those won’t be isolated pools of applications, deployed in separate environments and segregated from one another. Instead, applications will rely on connecting with one another, with the integrations spanning the two environments. This kind of complex interconnectivity will direct the architecture of cloud-based applications, and will direct the incremental development of legacy applications continuing to reside in on-prem environments. Succeeding as an enterprise IT organization in the future will require a strategy appropriate to this hybrid world, which is quite different than what many leaders currently envision their future will look like. 


The IDC chart presents a seeming paradox, one which we might call the Cloud Capex Conundrum. Public cloud providers, led by AMG, have been pouring money into capex as they build out global collections of massive data centers — and yet, all of that spend hasn’t seemed to affect ongoing IT capital spending by enterprises. This blog post has explained this as the result of a growing bifurcation of enterprise application portfolios: maintenance of the existing legacy applications, requiring ongoing investment in on-prem data centers, and a rapidly growing deployment of public cloud-based applications, resulting in massive capital spending by the public providers.

An implication of this bifurcation is the need for an enterprise computing strategy that incorporates architectures that can span the two environments, allowing applications to connect across the Internet in order to access other applications or data. The need for rich interconnection and what that requires is underappreciated by most enterprise IT leaders. I’ll discuss the effects this computing strategy will have on future enterprise engineering activities and organizations in my next blog post.



  1. […] my last blog post I did a lengthy analysis of “The Cloud Capex Conundrum” — the seeming paradox that on-prem capital investment is holding steady while public cloud […]

  2. Saad Tahir says:

    Interesting post Bernard. I think that the remaining on-prem enterprises are waiting out to learn for early adopter’s mistakes. They are also probably reviewing their tech stacks and breaking down monolithic applications and making them migration friendly. For the last on-prem stronghold to migrate, Cloud platforms will need to double down on something I call MAAS – Migration As A Service. Services like Anthos that help you migrate and manage your application anywhere. Migration is still a very manual intensive task that has a high level or risk due to unknown dependencies and deeply integrated systems.

  3. I think this bifurcation explains (and perhaps indicates a strong future for) on-premises service offerings such as Microsoft’s Azure Stack. Clearly, this is designed to address the needs of orgs to maintain whatever level of on-premises infrastructure is needed but without the issue of creating bespoke, difficult to maintain systems. This no doubt also explains the popularity of Kubernetes as an on-premises infrastructure modernization strategy which can span to public cloud.

    • Bernard Golden says:

      Dwayne, I completely agree. However, I suspect that most on-prem investment will be maintenance (i.e., keeping the current level of capability operational) and less will be transformational (i.e., making on-prem infrastructure as agile and available as public cloud capability). That’s not to say that there won’t be some organizations doing much more, and many organizations will do some, but I suspect that the percentage of transformed on-prem infrastructure will be 20% or less. I’ll be developing this theme and the implications in my next post.

      Thanks for your comment!

Leave a Reply

Your email address will not be published. Required fields are marked *