When the Cloud Doesn't Make Sense

I am a huge proponent of the cloud. I believe cloud services can help organizations move faster, improve resilience, and, when implemented well, strengthen security. But that does not mean every workload belongs there.

The better question is not, “Should we use the cloud?” It is, “Where should this workload run, and why?”

There are still situations where on-premises, edge, or hybrid infrastructure makes more sense. Here are a few that come to mind.

Massive Storage Used at One Site

Cloud storage is phenomenal when data needs to be shared broadly, replicated across regions, integrated with applications, or served to users all over the world. It is not always the best answer for large files used primarily inside one location.

Media libraries, local file shares, user mapped drives, surveillance archives, and similar workloads can become expensive or inefficient when every read and write has to cross the internet. If the people and systems using the data are in one building, local storage may be faster, simpler, and more cost-effective.

Latency-Sensitive Applications

If an application is highly sensitive to latency, it may be worth keeping it close to the people, devices, or equipment that depend on it. Manufacturing systems, local control systems, high-performance workstations, and some real-time applications can struggle when every transaction depends on a remote network path.

That does not mean the cloud should be dismissed immediately. There are strong cloud, edge, and hybrid options that can help. But latency-sensitive workloads deserve an honest architecture discussion. You still need resilience, backups, monitoring, and a recovery plan, and the cloud may be an excellent fallback or secondary location.

Homelab and Experimentation

This is my main personal use case for on-premises services. I have access to a lot of AWS services, but sometimes it is cheaper and more practical to break things in my own environment.

In a homelab, I can test routing ideas, build and tear down virtual machines, experiment with storage, and learn through trial and error without worrying about usage charges. For anyone interested in that path, I recommend taking a look at Proxmox. I started using it this year and love what it offers for homelab users and small businesses.

Certain Network Services

Some network services are naturally local. DNS filtering, VPN termination, DHCP, LAN monitoring, and other network-adjacent tools often make more sense close to the network they support.

Pushing all traffic to the cloud can introduce latency, complexity, and cost. In some cases it is absolutely worth it, especially for distributed organizations. In others, a local service is the cleaner answer.

The Real Answer Is Usually Hybrid

Maybe in a future post I will write about all the things I would put in the cloud, but that is a much longer list.

For now, the point is simple: the cloud is powerful, but placement matters. The right answer depends on latency, cost, security, resilience, data movement, operational skill, and the business outcome you are trying to support.

So, what do you want to run on-premises, what belongs in the cloud, and where does a hybrid approach make the most sense?