Self-host or cloud: the five-year math
We run our own mail, chat, monitoring and firewalls, so people assume we'll push self-hosting. We also migrate clients to Microsoft 365 every month. The honest answer to "cloud or on-prem" is a spreadsheet, and this is how we fill it in.
Start with the workload, not the ideology
The decision is per system, not per company. Email, identity and office documents have a strong gravitational pull toward Microsoft 365 or Google Workspace: the per-user price buys security tooling and availability a small shop can't replicate, and every vendor integrates with them. A line-of-business database with 30 users in one building, a file server with terabytes of CAD files, a camera system, or a lab environment often cost far less to run on a box in the closet — if someone competent looks after the box.
The costs people forget on the cloud side
- Egress and growth. Storage is cheap to put in and expensive to take out or move. Model what the data looks like in year five, not today.
- The licence tier you'll actually need. The cheap SKU rarely includes the security features you'll be told you need six months in. Price the tier with conditional access, device management and advanced threat protection.
- Backup of the cloud itself. See our note on this; budget for it.
- Dependence on the internet link. If the office circuit goes down and everything is in the cloud, everyone goes home. A second circuit or LTE failover is part of the cloud bill.
- Price changes you don't control. Per-user SaaS prices have gone one direction for a decade.
The costs people forget on the self-hosted side
- The second person. One engineer who knows the box is a single point of failure. Either pay for documentation and a backup provider, or pay in downtime when they're on holiday.
- Hardware refresh at year five, not year nine. Amortize it from day one.
- Power, cooling and a UPS that gets its battery replaced. Small numbers, but they're real and they compound with a server that runs 24×7.
- Patching discipline. Cloud providers patch their platform; nobody patches your hypervisor unless it's someone's job.
- Off-site disaster recovery. The cloud's biggest genuine advantage is that the building burning down is their problem. Self-hosting needs a tested answer to that question, and it costs money.
How the math usually lands
For organizations under about 25 people with ordinary office workloads, fully cloud wins on total cost and on risk, because the self-hosted "second person" cost dominates. Between 25 and 150, hybrid is common: identity, email and documents in 365, a compact virtualization host for the database, files and anything with heavy local I/O, and a cloud backup target for both. Above that, or with unusual data volumes or regulatory constraints, serious on-prem starts paying for itself again.
There are also non-financial reasons that legitimately tip it: data sovereignty, a vendor that only supports on-prem, or simply wanting to own the thing. Those are fine reasons, as long as they're stated and the cost of them is written down.
Our rule: run the five-year numbers for both, including the costs above, show both to the client, and recommend the lower one unless there's a reason we can write in one sentence.
If you'd like us to run that spreadsheet for your environment, it's part of every assessment we do, and the result is yours whichever way you decide to go.