When we started learning about the Cloud, I believe most of us had learned about Public and Private Cloud. How Cloud differs from the On-Prem.
Those who are new to this, below can be your starting definition.
Public Cloud — Infrastructure and services owned and operated by a third-party provider, delivered over the internet and shared across multiple tenants. You consume it, you don't own it.
Private Cloud — Cloud infrastructure dedicated exclusively to a single organization — either hosted on-premises or by a third party — not shared with others.
On-Premises — IT infrastructure physically located and managed within a company's own facilities. You own the hardware, software, and the responsibility.
[These are just "Definitions", and don't worry if doesn't make much sense now.]
We know these definitions — but one random Thursday, a question hit me: what does this actually mean? How did the Cloud originate, what technically qualifies something as Cloud what really separates these three models at a deeper level?
I did some reading and gathered all the information you will ever need to know about this in one place.
Before Cloud: The On-Premise Era
Before Cloud Computing was established, the entire IT world ran on only one model: On-Premise Infrastructure.
If a company needed computing power, applications, databases, or storage — it had to build everything itself.
A typical enterprise infrastructure stack looked something like this:
Physical Datacenter — Dedicated server rooms
Compute — Rack servers from vendors like Dell, HP, IBM
Networking — Routers, Firewalls
Licenses — Windows, VMware, Databases
These were hosted in the company's own premises and used to run applications such as ERPs or File Servers. Everything was local to the company.
Going deeper, the underlying architecture typically involved:
Bare-metal servers — Physical compute with no abstraction layer
Hypervisors (Type 1) — VMware ESXi, Microsoft Hyper-V running directly on hardware to carve out Virtual Machines
SAN/NAS Storage— Storage Area Networks or Network Attached Storage for shared block and file storage
VLAN-based networking — Manual network segmentation through managed switches
Manual provisioning — Every new server, IP, or storage volume required human intervention and lead time, often weeks
Everything was statically configured.
Benefits: Full Control, Localisation, Compliance
Limitations: Capital Investment, Procurement, Maintenance, Scaling
The Transition
For years this worked for companies. Then came the Internet boom of the 90s and 2000s. Now IT wasn't limited to big companies — even small startups needed servers and infrastructure.
But building traditional datacenters was expensive and complicated. Not everyone could buy servers, build server rooms, and maintain them.
Fun Read — The Y2K Problem: A fear that computer systems storing years as 2 digits would interpret the year 2000 as 1900, potentially causing global system failures.
At some point, Hosting Providers started to emerge. Companies began offering shared hosting. This was the bridge between traditional on-premise and modern cloud computing.
But the missing link was clear — you were renting someone else's datacenter, not consuming infrastructure as a service.
Then Came Cloud Computing
Cloud Computing is the delivery of computing resources — servers, storage, databases, networking, software — over the internet, on demand.
How It Started
Cloud Computing as a concept traces back to the 1960s when John McCarthy suggested that computing would one day be organized as a public utility. The term "cloud" itself was used in network diagrams to represent the public network.
The real origin points:
1999 — Salesforce became the first company to deliver enterprise software over the internet, establishing the SaaS model
2002 — Amazon Web Services launched as a set of internal infrastructure tools
2006 — AWS launched EC2 (Elastic Compute Cloud) — pay-per-use virtual machines on demand
That 2006 moment is when Cloud Computing as an industry was truly born.
https://aws.amazon.com/about-aws/
What Is Cloud, Really?
Cloud is the computing that someone else builds and maintains the infrastructure; you consume it instantly, scale it up or down, and pay only for what you use.
Cloud became possible because of virtualization.
AWS uses a custom hypervisor called Nitro, which offloads virtualization to dedicated hardware. AWS previously used Xen before building Nitro.
Azure uses a modified version of Hyper-V, Microsoft's own Type-1 hypervisor.
I have an in-depth blog on Virtualization
What Qualifies as Cloud? The NIST Definition
NIST — the official standards body — defines Cloud by 5 essential properties:
On-demand self-service: User can provision resources automatically without human interaction from the provider
Broad network access: Capabilities available over the network, accessible from any standard device — phones, laptops, workstations
Resource pooling: Provider's resources are pooled to serve multiple consumers (multi-tenancy), dynamically assigned based on demand
Rapid elasticity: Resources can be scaled up or down quickly — sometimes automatically — to match workload demand
Measured service: Usage is monitored, controlled, and billed — you pay for what you consume
For something to qualify as Cloud, it must satisfy all five of these properties.
Not every provider that rents you a server qualifies as Cloud. A traditional hosting provider gives you a fixed server — static, manually managed, not elastic. Shared hosting fails on elasticity, on-demand self-service, and measured service. So be precise when using the word "Cloud".
Cloud = Infrastructure as APIs + Software
What truly separates Cloud from On-Prem isn't just where the hardware lives — it's the software layer on top. Cloud infrastructure is fully programmable. Every resource — a VM, a database, a network — is created, configured, and destroyed through an API call. That software layer is what transforms raw hardware into a Cloud.
This becomes the differentiator between the Private Cloud and On-Prem
Public Cloud
Public Cloud is a multi-tenant environment where infrastructure is owned and operated by a Cloud Service Provider (CSP) and shared across thousands of customers, with strict isolation enforced at the software and hypervisor level.
Key technical characteristics — and a few things we often overlook:
Blast radius isolation — Your workload runs on shared physical hardware, but memory, CPU, and network are isolated via the hypervisor. AWS Nitro enforces this at the hardware level, not just software.
The Shared Responsibility Model — The CSP secures the infrastructure; you secure what runs on it. This is a contractual and architectural boundary, not optional.
Availability Zones (AZs) — Physically separate datacenters within a region, connected by low-latency private fiber. Designing across AZs is not automatic — it is an architect's deliberate decision.
Egress costs — Data coming IN to public cloud is free. Data going OUT is charged. This is one of the most underestimated cost drivers in public cloud.
Private Cloud
A Private Cloud is a cloud environment — satisfying all five NIST properties — dedicated exclusively to a single organization.
On-Premises Hardware + Management/Orchestration Software = Private Cloud
The management layer — OpenStack, VMware vSphere, Nutanix Cloud Platform — is what turns a datacenter into a Private Cloud. Without it, you just have servers.
Single-tenant by design — No shared compute with external parties. Full control over data residency.
Cloud-like experience — Self-service portals, automated provisioning, elastic scaling within owned capacity, API-driven infrastructure.
Things engineers often don't know about Private Cloud:
A Private Cloud has a hard capacity ceiling — you can only scale to what you physically own. Elasticity is bounded, unlike Public Cloud.
Hosted Private Cloud exists — providers like IBM Cloud or Rackspace can run a dedicated private cloud on their hardware, for you. It is still private cloud, just not on your premises.
OpenStack is the dominant open-source platform for building private clouds and powers many telco and government private clouds globally.
https://www.openstack.org/
Cloud isn't a place. It's a model. Now you know what that model actually means.