On-prem is short for on premises. The organization's IT infrastructure is physically located within the company's own facilities. The company owns the servers, the storage arrays, and the networking gear, and that hardware sits in a data center inside one of their buildings. From the 1960s through the early 2000s, this was the only viable option if you needed to run applications or store data at any meaningful scale.
Owning the hardware gave you full control. It also handed you every cost and responsibility that comes with it. You paid capital expenditure for the servers themselves, but also for the data center around them, climate control, physical security, power, and the IT team that handled upgrades, patches, and repairs. None of that work disappears once the servers are racked. It is ongoing.
Five characteristics of the on-prem model show up repeatedly in PCA questions, and they are worth committing to memory.
Full control. Companies own and manage the hardware end to end. Nothing leaves the building.
High costs. Significant capital expenditure for servers, data centers, climate control, and physical security. This is capex, not opex, and it lands before you serve a single user.
Scalability challenges. Scaling required purchasing and installing new hardware, which took time and money. If demand spiked next quarter, you were ordering servers now and hoping they arrived in time.
Maintenance requirements. Dedicated IT teams handled upgrades, patches, and repairs. The hardware does not maintain itself, and the people who maintain it are payroll.
Underutilization. Infrastructure often sat idle to handle peak usage but was underused most of the time. You sized for the worst day of the year and paid for that capacity every other day.
These five together explain why companies eventually went looking for alternatives.
With cloud computing, the organization's IT infrastructure is hosted offsite, by another company, and is accessed through the internet. Instead of owning the hardware, you pay a cloud service provider, GCP, AWS, or Azure, for the services you actually use. The provider handles the physical hardware, the maintenance, and the scalability. You focus on running your workloads.
This shift in ownership is why we call what users access through the cloud "cloud services." You are not buying servers. You are renting capabilities. The cloud service provider, in our case Google Cloud, provides services to the user or organization consuming their resources over the internet.
For Professional Cloud Architect questions, the on-prem versus cloud contrast usually shows up as a tradeoff:
When a PCA scenario describes a customer with aging hardware, long procurement cycles, or capacity that sits idle most of the year, those are the cues that the question is testing whether you understand the cloud model's answer to each of those problems.
My Professional Cloud Architect course covers on-premises versus cloud alongside the rest of the foundational architecture material.
