On-premises vs Cloud
😟 THE PROBLEM
The server room is full, hot and out of warranty next year.
😊 THE SOLUTION
Move suitable systems to the cloud and stop buying hardware.
- A hardware refresh costs millions✓ No hardware to buy in the cloud
- Power, cooling and repairs never stop✓ Microsoft handles the datacenter
- Capacity is stuck at what you bought✓ Add or remove capacity any time
Quick answer
On-premises means you buy, house and run your own servers in your own building or leased datacenter, so you control everything and are responsible for everything. In the cloud you rent computing resources from a provider over the internet. The provider runs the hardware, you pay for what you use, and capacity changes in minutes.
Exam objective: 1.1.1 Define cloud computing. 1.1.4 Identify appropriate use cases for each cloud model. (Skills measured as of July 20, 2026)
Start with a question
Would you rather own a car or use a taxi app?
Owning a car means you choose the model, decide who drives it and leave your things in the boot. You also pay for it up front, insure it, service it, park it and fix it when it breaks, even in the months you hardly drive.
A taxi app means no purchase, no servicing, and a car whenever you need one. You give up some control and pay per ride.
On-premises IT is owning the car. The cloud is the taxi app. Neither is “right” for everyone. It depends on how you travel.
Explain it like I’m 10
The technical explanation
On-premises (often shortened to on-prem) means an organisation runs IT infrastructure that it owns, in a server room or datacenter it controls. It is responsible for the whole stack, from the building and electricity up to the applications.
Cloud computing means renting infrastructure and services from a provider over the internet. The provider owns and operates the datacenters, hardware and core networking. The customer configures and uses the services.
| On-premises | Cloud (Azure) | |
|---|---|---|
| Up-front cost | High: buy servers, storage, network gear | None or low: pay as you go |
| Spending model | Mostly CapEx | Mostly OpEx |
| Time to get a server | Weeks (order, ship, install) | Minutes |
| Capacity | Fixed until you buy more | Scale up or down on demand |
| Physical maintenance | Your team | The provider |
| Physical control | Full | None (you control configuration) |
| Global reach | Build or lease sites yourself | Deploy to regions worldwide |
| Disaster recovery | Build and pay for a second site | Use other zones and regions |
| Security | Entirely your responsibility | Shared with the provider |
Architect level Comparing costs honestly: total cost of ownership (TCO)
A fair comparison includes everything a workload costs over 3–5 years:
- On-premises: hardware (with refresh), software licences, datacenter space or lease, power and cooling, network circuits, backup and DR site, support contracts, and staff time to operate it.
- Cloud: compute, storage, networking (including outbound data transfer), licences (or Azure Hybrid Benefit to reuse existing Windows Server and SQL Server licences), support plan, and the staff time still needed to operate it.
Two traps catch experienced teams. They compare an optimised on-prem estate with an unoptimised lift-and-shift, and they forget egress and cross-region data transfer. Microsoft’s pricing calculator estimates cloud costs. Azure Migrate’s business case feature compares on-prem and Azure TCO from real discovery data.
Real-world business scenario
A regional bank
- The problem
- The core banking system must stay in the bank's own datacenter for regulatory and integration reasons, but its mobile app and analytics needs are growing fast and the datacenter is full.
- What they used
- Keep the core system on-premises; run the mobile API on Azure App Service and analytics on Azure, connected privately with ExpressRoute.
- Why that choice
- Regulated core stays under direct control; customer-facing and bursty analytics workloads get cloud scale without expanding the datacenter.
- The result
- New mobile features ship faster, month-end analytics finish in hours instead of days, and the datacenter expansion is cancelled.
This is very common. The choice is rarely “all on-prem” or “all cloud”. It is about putting each workload where it fits best, which leads straight to hybrid cloud.
When to use it and when not to
✓ Use it when…
- **Choose cloud** for new applications, variable demand, global users, and when you want managed services.
- **Choose cloud** when a hardware refresh, lease renewal or datacenter capacity limit is approaching.
- **Stay on-premises** for workloads tied to local hardware or very low latency on site.
- **Stay on-premises** where regulation or contracts genuinely require it.
✕ Think twice when…
- Moving to the cloud without changing anything, then expecting big savings.
- Keeping everything on-premises just because "that is how we have always done it".
- Ignoring skills. Both models need trained people, just different skills.
↔ Alternatives to consider
- **Hybrid cloud**: combine both and connect them securely.
- **Azure Local** (formerly Azure Stack HCI): run Azure-managed infrastructure in your own datacenter.
- **Azure Arc**: manage on-premises servers from Azure with the same tools and policies.
Interview preparation
Say your answer out loud first, then open the model answer. Interviewers care more about your reasoning than the exact words.
Beginner What is the difference between on-premises and cloud?
Model answer
On-premises means the company owns and runs its own servers, so it handles purchasing, power, cooling, security and repairs. Cloud means renting those resources from a provider like Azure over the internet. The provider runs the physical side and the company pays for what it uses.
Interviewer follows up
Which one gives more control?
You
On-premises gives full control over hardware and the physical environment. The cloud trades some of that low-level control for speed, scale and less maintenance, while you still control your configuration, data and access.
Intermediate Name the hidden costs of on-premises infrastructure that people often forget.
Model answer
Power and cooling, floor space or datacenter leases, physical security, hardware support contracts, spare parts, the hardware refresh every few years, backup and disaster-recovery sites, software licences and, above all, staff time to run it. A server's purchase price is usually a small part of its total cost of ownership.
Experienced developer When is staying on-premises the right architectural decision?
Model answer
When latency to on-site equipment is critical (manufacturing control systems), when data sovereignty or contractual rules cannot be met by any available region, when there is a large, steady workload on recently purchased hardware, or when the application is tightly coupled to mainframes or hardware that cannot move. Even then, I would look at hybrid patterns, managing on-premises servers with Azure Arc or running Azure services on-premises with Azure Local, to get cloud-style management.
Scenario A company's server room lease ends in nine months, and its servers are five years old. What would you recommend?
Model answer
This is a classic trigger for migration. I would assess workloads with Azure Migrate, move the straightforward ones (lift and shift to VMs, or replatform to PaaS where cheap to do), retire unused systems, and keep anything that truly must stay on-premises in a smaller hybrid footprint. That avoids a costly hardware refresh and a new lease.
Test yourself with exercises
4 original practice questions in the AZ-900 style. Pick your answers, then check them. You will see why each option is right or wrong. Score 75% or more and this topic is marked complete.
Quick revision
Key points
- On-premises = you own the hardware, the building concerns and all of the maintenance.
- Cloud = you rent capacity; the provider runs the physical infrastructure.
- On-prem is CapEx-heavy with fixed capacity; cloud is OpEx with on-demand capacity.
- On-prem gives maximum physical control; cloud gives speed, scale and global reach.
- Many organisations run a mix of both (hybrid).
Important terminology
- On-premises (on-prem)
- IT infrastructure owned and operated by the organisation itself, in its own building or a leased facility.
- Datacenter
- A building that houses servers, storage and networking with power, cooling and physical security.
- Hardware refresh
- Replacing ageing servers, typically every 3–5 years, which is a major recurring capital cost on-premises.
- Lift and shift
- Moving an application to the cloud with as few changes as possible, usually onto virtual machines.
Common mistakes
- Comparing only server purchase price with a monthly cloud bill. Compare total cost of ownership.
- Assuming on-premises is automatically more secure. Security depends on how well it is run.
Exam tips
- If a scenario stresses "complete control over physical hardware", the answer is usually on-premises or private cloud.
- If it stresses "no up-front cost" or "scale quickly", the answer is usually cloud.
Five quick revision questions
-
1.Who maintains the physical servers on-premises?Show
The organisation itself.
-
2.What is a hardware refresh?Show
Replacing ageing servers every few years, a recurring capital cost on-premises.
-
3.Which model gives more physical control?Show
On-premises.
-
4.What does "lift and shift" mean?Show
Moving an application to the cloud with minimal changes, usually onto VMs.
-
5.Which spending type dominates on-premises - CapEx or OpEx?Show
CapEx (capital expenditure).
Official references
Finished this lesson? Tick it off. Your progress is saved in this browser.
Next up: Public Cloud