The cloud question gets argued as though it were a philosophy. Cloud-first. Cloud-native. Or, from the other camp, nobody else touches my data. Both positions make good conference talks and terrible purchasing decisions, because neither asks the only question that matters: what is this workload doing, and where does it run best?
The real answer is boring, and it has been boring for a while. Almost every business that has been running for more than a few years ends up hybrid, meaning some things live in the cloud and some things stay on equipment you own. Not because hybrid is elegant, but because that is what you get when you decide workload by workload instead of by ideology. Harrison puts it more bluntly: the cloud is just someone else’s computer. Sometimes someone else’s computer is a much better choice than yours. Sometimes it is not.
Hybrid Won, and Nobody Threw a Party
Flexera’s 2026 State of the Cloud Report found that 73 percent of the organizations it surveyed use hybrid cloud, combining public and private cloud environments, and 33 percent use multiple public and multiple private clouds at once. That is not a transitional state on the way somewhere cleaner. That is the destination.
The reason is simple. Businesses do not migrate everything at once. They move email, then file sharing, then a line-of-business application, and along the way discover that one system talks to a machine on the shop floor and another produces files too large to be comfortable over an internet connection. Those stay. Ten years later you have a hybrid environment nobody designed, and it works fine as long as somebody manages both halves deliberately.
What Genuinely Belongs in the Cloud
Some workloads are so clearly better hosted that arguing about it is a waste of a meeting.
- Email. Running your own mail server in a small business stopped making sense years ago. Spam filtering, deliverability, mobile access, and security updates are constant work, and hosted platforms do them at a scale you cannot match.
- Collaboration and document sharing. Anything multiple people edit, comment on, or need version history for. The cloud is genuinely better at simultaneous access, and the alternative is emailing attachments named “final_v3_REAL.xlsx”.
- A copy of your backups. Not necessarily the only copy, but definitely one of them. Backups that live in the same building as the thing they protect are one flood, fire, or ransomware event away from being useless.
- Anything remote or mobile workers touch daily. If people need it from a truck, a client site, or a kitchen table, hosting it locally routes everything through your office connection, which becomes a single point of failure with a modest upload speed.
- Systems with spiky demand. If you need ten times the capacity for two weeks a year, renting beats owning. The classic cloud argument, still true where it applies.
What Legitimately Stays Local
Keeping something on-premises is not nostalgia. For these categories it is the correct engineering answer.
- Large files under heavy daily access. Design firms, video production, engineering, survey work, medical imaging. When your team opens multi-gigabyte files dozens of times a day, local network speed dwarfs your internet speed, and no syncing software makes that gap disappear.
- Anything tied to physical equipment. Machine controllers, lab instruments, point-of-sale hardware, building systems, cameras. These often need a specific computer on the same network, sometimes running an old version of software the manufacturer will never update.
- Low-latency line-of-business applications. Older business software was often designed to sit right next to its database. Move the database to the cloud and the application becomes technically functional and practically unusable, because every click waits on a round trip.
- Workloads with a genuine regulatory or contractual reason. Genuine, not assumed. Most compliance frameworks permit cloud hosting with the right controls, so verify the requirement before designing around it.
- Anything that has to keep working when the internet does not. Big providers do have bad days. Amazon Web Services published a summary of a disruption in its Northern Virginia region on October 19 and 20, 2025, in which knock-on effects on some services ran roughly fourteen hours. If your business cannot pause that long, something local needs to carry it. We looked at that in the cloud can go down, and what that means for your business.
The Cost Comparison People Get Wrong
The usual comparison is a server purchase price against a monthly cloud bill. That comparison is wrong in both directions, which is why both camps think they win it.
On-premises costs are systematically understated. The server is the smallest line. Add the backup target, the power protection, the network gear, the licensing, the electricity, the space, the replacement cycle, and the hours somebody spends patching and monitoring it. Divide across five years. That is your real monthly number, and it is not the one on the purchase order.
Cloud costs are understated too, differently. The subscription is predictable, the consumption is not. Storage that grows, data transfer charges, per-user licenses that outlive the users, backup retention, and the extra tools you buy when the first one does not quite cover it. Flexera’s 2026 State of the Cloud Report found 29 percent of cloud spend was wasted, the first increase in five years, which shows how easily this drifts even at organizations watching it full time. The same report found 43 percent of respondents cited comparing on-premises versus cloud costs as a migration challenge. Everyone finds it hard.
Do it properly: a five-year total for each option, including labor on both sides and the cost of your team’s time. That last part is where cloud often wins even when it looks more expensive on paper, an argument we made in why saving time takes priority over saving money.
The Exit Question Nobody Asks Until It Is Late
Before you commit a workload anywhere, ask one question: if we wanted to leave in three years, what would that take? Not whether you plan to. Whether you could.
Ask where your data lives, what format you could export it in, whether the export includes structure and history or only raw records, and what it would cost to move it out. Transfer charges used to make that last part painful. That has improved under regulatory pressure: Network World reported in March 2024 that Amazon Web Services would waive data transfer out charges for customers moving off its platform, following a similar move by Google Cloud in January 2024, with AWS pointing to the direction set by the European Data Act. Still, the practical cost of leaving is rarely the transfer fee. It is the rebuilding, the retraining, and the integrations written for one platform’s way of doing things.
You do not need a perfect exit plan for every system. You need to know roughly which platforms would take a weekend to leave and which would take six months. That changes how much you are willing to build on each one.
The Bottom Line
There is no cloud strategy. There is a list of workloads, and each has a right answer based on how it is used, how much data it moves, what it connects to, and what leaving would cost. Most businesses land on hybrid because most have a mix of those situations. That is not indecision, it is accuracy.
The practical starting point is a list. Write down every system your business depends on and put one word next to each: cloud, local, or undecided. The undecided ones are where the actual conversation lives, and there are usually fewer of them than people expect.
Harrison Ward Technology helps small and mid-sized businesses in Denton County work through that list without an agenda about the answer. Sometimes we recommend moving something to the cloud. Sometimes we recommend a server in your closet, because that is what the workload needs. If you want a workload-by-workload assessment rather than a migration pitch, we are glad to go through it with you. Contact us today
Sources:
Comments are closed