How big is the Kubernetes market? Here’s an estimate from a recent IDC PDF:
The rapid growth of cloud-native applications, driven by DevOps, containers, and microservices, has made container orchestration a critical capability because containers are challenging to manage at scale without an orchestration layer. This trend is driving significant market growth, with IDC forecasting the container infrastructure software [this is what IDC calls the Kubernetes market] market to reach $6.4 billion by 2028. The focus on infrastructure optimization is a key initiative, with 48% of organizations identifying it as a top priority for developing AI applications.
This is a tiny market, and it’s not even 2028 yet. Why “tiny”? Think through slicing up that market between Google, Azure, AWS, VMware, Oracle/OCI, whatever other big company I’m thinking of, and then all the startups. Of course, maybe you don’t pay for Kubernetes directly, and you just pay for the compute it uses.
If you have access to the original market sizing report, you can dig into that $6.4bn estimate for 2028. The estimate is that 58.1% is on-premises, and 41.9% in public cloud. You also see that the estimate for the market size in 2025 is $3.72bn. It’s a good report, you should check it out if you can.
For reference, if I’m understanding it, IDC says the Linux market has a “projected spending set to grow from $93.0 billion in 2023 to $192.1 billion in 2028 at a CAGR of 15.6%.” Yes, that’s a very mature market. Mainframes might be the only OS market more mature! Point being, comparing a ten year old market to a ~35 year old market isn’t exactly apples to apples.
More importantly, spend doesn’t mean everything for importance and usage. The best way to judge the value of a platform is how many applications run on it globally. Those numbers are hard to find. Last I looked, the wildly savings hands consensus is something like 10% to 20% of apps globally run on Kubernetes. I suspect that it’s much closer to 10% in “enterprises.”
People really want to use Kubernetes, Kubernetes is still hard to use
More from the report:
Kubernetes has become the de facto standard, often called the “operating system of the cloud” because it orchestrates containerized applications across a cluster of nodes. The demand for containers and Kubernetes is accelerating, particularly with the rise of generative AI (GenAI) coding assistants, which are speeding up the development and deployment of AI-driven applications. Organizations are increasingly choosing to build their AI applications on Kubernetes while seeking the security of a private cloud to keep AI workloads close to their sensitive corporate data.
Yes, but:
Unfortunately, Kubernetes introduces a significant level of complexity, making it challenging to deploy, configure, and manage, especially at scale. This complexity stems from various abstraction layers and intricate requirements for networking and security. A significant source of this difficulty is its reliance on lengthy and complex YAML configuration files. Maintaining these files is error prone, requiring meticulous attention to detail. Managing all the files, parameters, and settings across different resources becomes a demanding task that can easily overwhelm developers.
Getting value out of Kubernetes requires a platform
Kubernetes is still so complex you need to wrap a platform around it. If you use Kubernetes, you’re going to end-up building a platform for it. Building your own platform, especially on Kubernetes, will take years and millions of dollars, and the result is often not good.
Building your own platform is costly and risky. Many people who build out this platform end up with something that’s not much better. And, when you’re successful, you know have 10+ year responsibility to keep building out that platform. As you build it more, customizing to what you need, you fork more and more from industry norms and have your own special platform. Now, there’s no help for you. You’ve built your new mainframe. You’ve built your own lock-in and killed any hopes of portability.
This is fine for the likes of Google, Facebook, etc. who have the means to staff people to build those platforms. But, even the biggest banks and governments should look at the spend they have on building and maintaining their own platform and ask “is this the best way to spend our time and money?” No, it’s not.
As one of the Kubernetes OGs said recently:
Do not blindly start with Kubernetes. Seriously. If your application can get by with a simple PaaS or Serverless offering I’d consider that first. Even VMs make sense for most situations.
Instead of “start with Kubernetes”, I’d recommend people “decouple from the underlying platform”, which is the number one reason to package your application using a container image, which runs on most PaaS offerings, VMs, and bare metal.
The Kubernetes people themselves are literally telling you not to build out Kubernetes unless you really, really have to.
The platform engineering community grew out of this need: they’re the ones who are trying to make Kubernetes useful by putting a platform on-top of it. From the PDF: “The increased adoption of platform engineering has been primarily fueled by the need to manage Kubernetes across the enterprise.” We’re having to invent a whole new methodology and set of tools to make Kubernetes work as a platform, and most of those tools layer on-top of and hide Kubernetes from the operations and developers that run their apps on Kubernetes.
The dead DevOps people are probably spinning joyfully in their graves, trying to yell “how you like it now?" muffled through the six feet of dirt the IDP people poured on them a few years ago.
The 10 year distraction
We of course solved all of this back in the PaaS days. Back then we called it “platform as a product.” And, there’s still many organizations running PaaSes that are doing just fine.
Over the past 8 or so years, all of us - buyers and sellers in the tech industry - have been distracted by Kubernetes and lost sight of what real PaaSes look like and how well they work. You know, for huge, massive, real-world companies running core apps at scale, for years on end.
If you want to see a winsome look back at the pre-Kubernetes days where things just worked, check out Paula’s and Derik’s talk from Cloud Foundry Day.
Anyhow.
It’s been ten years. People still say Kubernetes is too complex. It’s a tiny market. You need to build a platform on-top of it to make it useful. Founding Kubernetes Kommunity people are saying you should think real hard before using it.
¯_(ツ)_/¯
P.S.: Need a real platform? May I suggest: TryTanzu.ai.