Blog
Do You Actually Need Kubernetes and the Cloud?
Kubernetes is everywhere.
According to the CNCF Annual Cloud Native Survey, 98% of surveyed organisations have adopted cloud-native techniques, while 82% of container users run Kubernetes in production. The survey also found that 66% of organisations hosting generative AI models use Kubernetes for some or all inference workloads.
Cloud-native infrastructure has clearly gone mainstream.
But that doesn’t mean every application needs Kubernetes.
So, Do You Actually Need Kubernetes?
It depends on the workload, your team and the level of complexity you actually need.
Kubernetes is designed to orchestrate and manage containerised applications at scale. It can be particularly useful when you have:
- Multiple applications or services
- Complex scaling requirements
- Several development teams
- A need for standardised deployment and operations
- Hybrid or multi-cloud environments
- Large-scale AI or data workloads
AWS recommends evaluating compute choices on a workload-by-workload basis, taking organisational structure, team skills and operational requirements into account. Its guidance positions Kubernetes as a strong fit for organisations investing in platform engineering, while serverless options can reduce infrastructure-management overhead.
When Might Kubernetes Be Overkill?
Not every application needs a cluster.
If you’re running a relatively simple application, API or internal tool, you may be better served by:
- A managed container platform
- Serverless infrastructure
- A virtual machine
- A traditional application and database
Google Cloud’s architecture guidance explicitly recommends considering serverless first when a workload is a good fit, noting that Kubernetes adds complexity and should be used when its capabilities are actually needed.
That doesn’t make Kubernetes bad.
It just means:
| “Powerful technology is only valuable when you actually need the power.”
Kubernetes vs Serverless
This is often less about technology and more about operational responsibility.
Serverless can be a good fit when you want to focus primarily on building applications while the cloud provider manages more of the underlying infrastructure.
Kubernetes gives you more flexibility and control, but usually requires greater operational maturity.
AWS’s current guidance highlights this trade-off clearly: many workloads can work successfully with either model, and the decision often depends on your team’s skills, organisational structure and need for standardisation.
What About AI?
AI is making Kubernetes increasingly relevant — particularly for organisations running their own models or managing inference workloads at scale.
The CNCF’s 2026 survey found that 66% of organisations hosting generative AI models use Kubernetes for some or all inference workloads.
But here’s the important distinction:
Using AI does not automatically mean you need Kubernetes.
An application making API calls to an external LLM has very different infrastructure requirements from a company running its own models, GPUs and distributed inference systems.
Again:
Cloud or On-Premises?
Cloud is not automatically better.
And it is not automatically cheaper.
The right decision depends on factors such as:
- Workload requirements
- Security and compliance
- Data location
- Existing infrastructure
- Team expertise
- Scaling needs
- Operational costs
For some organisations, cloud provides flexibility and access to managed services that would be difficult to build internally.
For others, particularly with predictable workloads or existing infrastructure, on-premises or hybrid environments may make more sense.
The Questions You Should Actually Ask
Before choosing Kubernetes, serverless or a cloud platform, ask:
How complex is the application?
How many services do we actually need?
How much scaling do we realistically expect?
Do we have the skills to operate Kubernetes?
How much infrastructure do we want to manage ourselves?
Are we solving a real requirement — or following a trend?
Because sometimes the answer is Kubernetes.
Sometimes it’s serverless.
Sometimes it’s a managed container platform.
And sometimes…
it’s just an application, a database and a deploy button.
The Bottom Line
Kubernetes is no longer an experimental technology. It has become a major part of modern production infrastructure and is increasingly important for large-scale AI workloads.
But popularity is not a technical requirement.
The best infrastructure isn’t the most complicated one.
It’s the one that gives your team what it needs — without creating unnecessary complexity.
At EPSza, we believe good software engineering starts with understanding the problem before choosing the technology.
Because sometimes you need Kubernetes.
And sometimes you really just need to ship the app.