Outcome
By the end of this lesson, you will be able to compare owning computers with renting them from the cloud, and say which costs move from up-front to monthly.
| Prerequisites | L1 |
| Time | 8 minutes |
1. Concept
Before the cloud, almost every company ran its own computers. This is called on-premises, usually shortened to on-prem: the machines sit on your premises, or in a building you lease space in, and you own them.
Many companies still run on-prem, and many run a mix of both. Neither is wrong. They are two different ways of paying for the same thing.
What owning involves
If you own your computers, you pay for:
- The hardware. Servers, disks and network equipment, bought up front.
- The room. Floor space, racks (metal frames that hold servers), cooling and backup power.
- The people. Staff who install machines, replace broken parts and apply updates.
- Spare capacity. You must buy enough for your busiest day, even if that day comes once a year.
Most of that is paid before you serve a single customer. Accountants call this capital expenditure, or capex: money spent up front on things you keep for years.
What renting changes
In the cloud, the provider owns the hardware, the building and the repair staff. You pay a monthly bill based on what you used. Accountants call this operating expenditure, or opex: money spent month by month as you go.
That shift has three consequences.
- Capacity planning gets easier. You no longer have to guess next year’s peak and buy for it. You rent more on the busy day and less on the quiet one.
- The work moves, it does not vanish. The provider looks after the physical machines. You still decide what to rent, set it up, keep it secure and switch it off.
- Cost becomes variable. The bill changes every month with usage. That is flexible, and it is also harder to predict.
Who is responsible for what
A simple way to picture it: think of the cloud as a building with layers. The provider is responsible for the bottom layers (the building, power, physical machines and the network between them). You are responsible for what you put on top (your software, your data, who can access it, and how much of it you leave running).
Providers call this the shared responsibility model. The detail differs by service, but the principle is the same everywhere: renting the machine does not mean handing over the decisions about it.
When owning can still make sense
Owning can cost less when the workload is large, steady and predictable for years, and you already have the buildings and staff. Some companies also keep data on-prem for legal or regulatory reasons. The cloud wins most clearly when demand changes, when speed matters, or when you would rather not run a data centre at all.
2. Worked example
A team needs enough computing power for a busy season that lasts two months a year. For the other ten months it needs about a quarter of that.
All numbers are illustrative.
Owning. The team must buy enough servers for the peak: say 8 servers at $4,000 each, so $32,000 up front, plus an illustrative $500 a month for space and power. Over three years:
| Item | Cost |
|---|---|
| 8 servers, bought once | $32,000 |
| Space and power, 36 months at $500 | $18,000 |
| Three-year total | $50,000 |
For ten months of every year, six of those eight servers sit mostly unused.
Renting. The team rents 8 machines during the two busy months and 2 machines for the other ten, at an illustrative $150 per machine per month.
| Item | Cost per year |
|---|---|
| Busy months: 8 machines x 2 months x $150 | $2,400 |
| Quiet months: 2 machines x 10 months x $150 | $3,000 |
| Per year | $5,400 |
| Three-year total | $16,200 |
The rented version is cheaper here because demand goes up and down. If the team needed all 8 machines all year, renting would be 8 x 12 x 3 x $150 = $43,200. That is closer to owning, before counting the staff the owner also needs. The shape of demand decides which option wins.
3. Hands-on (5 min: no account needed)
1. Open the AWS Pricing Calculator: calculator.aws2. Choose "Create estimate", then add Amazon EC2.3. Pick any region and a small machine size (for example 2 vCPUs, 8 GiB memory).4. Set usage to 730 hours a month. Note the monthly cost.5. Change usage to 160 hours a month (about 8 hours a day on weekdays). Note the new cost.The second number is what you would pay if the machine were only on during working hours. Owning hardware gives you no such choice: you paid for it whether it runs or not.
You do not need to save or sign in to anything to use the calculator.
4. Knowledge check
Q1
What does “on-prem” mean?
A. Software that runs only in a web browser
B. A premium tier of cloud service with extra support
C. Computers a company owns and runs in its own space
D. A cloud machine that is switched off overnight
Show answer
Correct: C. On-premises means you own and run the hardware. B sounds plausible because of the word “premium”, but it is unrelated. A and D describe other things entirely.
Q2
Moving from owning servers to renting them in the cloud mostly changes cost from:
A. Monthly payments to one large payment up front
B. Up-front spending on equipment to monthly spending on use
C. Variable spending to a fixed amount that never changes
D. Paying for hardware to paying for nothing at all
Show answer
Correct: B. Owning means capital spent up front (capex). Renting means operating spend month by month (opex). A and C are the wrong way round. D is wrong because you still pay; you just pay as you go.
Q3
Under the shared responsibility model, which job stays with you when you rent cloud machines?
A. Deciding what to rent and switching off what you no longer need
B. Replacing failed disks inside the provider’s data centre
C. Paying the electricity bill for the provider’s building
D. Hiring security guards for the provider’s data centre
Show answer
Correct: A. The provider looks after the building, power and physical machines. What you rent, how it is set up and how long you leave it running remain your decisions, and your cost.
5. Apply
When someone says “we should move to the cloud to save money”, ask one question first: does our demand go up and down, or is it flat all year? Renting saves the most when demand changes and when unused machines are actually switched off. If neither is true, the saving may be smaller than people expect.
Related lessons
- L1, Renting someone else’s computers by the hour
- L3, Meet the big three: AWS, Google Cloud and Azure (next)