Platform-as-a-Service
Platform-as-a-Service (PaaS) means: a provider makes a ready-made environment available over the internet in which one can run one's own programs. So you're not just renting computers, but a complete working environment including maintenance.
Anyone who wants to put a program on the internet needs more than just the program code. It needs a computer that runs day and night. An operating system must be installed on this computer, meaning the basic software such as Windows or Linux. Then there are helper programs, security updates, and a database in which the users' data is stored. Platform-as-a-Service means: a provider takes care of all of that and rents it out as a ready-made service. You just upload your own program code; the rest is already prepared.
Why developers want to avoid server maintenance
Operating servers is work that has nothing to do with the actual product. Someone has to install updates, monitor hard drives, and respond to attacks. In a small team, this can quickly tie up an entire person. At a PaaS provider, that provider’s staff takes care of it, spread across thousands of customers at the same time.
The second advantage is speed. An idea can be online within an afternoon instead of after weeks of preparation. For start-ups, this is often decisive, because they need to quickly find out whether anyone even wants their product at all. Many well-known apps ran entirely on rented platforms during their first years.
For this, you pay a price, in two forms. First, PaaS is more expensive per unit of computing power than a bare rented server. Second, you tie yourself to the provider’s rules. If you later switch to another one, part of the program often has to be rewritten. Experts call this dependency lock-in.
What the provider takes on and what it doesn’t
Cloud offerings can be understood as a ladder of levels. At the lowest level, you only rent virtual machines and take care of everything on top yourself; this is Infrastructure-as-a-Service. At the top level, you use ready-made programs in the browser, such as an online word processor; this is Software-as-a-Service. PaaS sits exactly in between: the underlying structure is ready-made, and you write the application yourself.
Technically, this usually works as follows: you connect your code repository to the platform. With every change, the platform automatically rebuilds the program and starts it. If the load changes because many users suddenly show up, it starts additional copies of the program. This automatic expansion is called scaling and is one of the main arguments for PaaS.
A common misconception is that PaaS takes over all the work. You still have to write, test, and maintain the program itself. No platform catches errors in your own code either. It only ensures that everything underneath runs stably.
PaaS in AI projects and in the business press
Well-known examples are Heroku, Google App Engine, Microsoft Azure App Service, and Vercel. Anyone using a website or an app doesn’t notice any of this. The difference is invisible to the user and only concerns the company behind it.
In the AI field, a distinct variant has developed. Providers offer ready-made environments in which you can train or run models, including graphics cards and matching software. Since individual graphics cards for AI cost tens of thousands of euros, even large companies prefer to rent them by the hour.
In business news, PaaS usually comes up in the quarterly figures of Microsoft, Amazon, and Google. Their cloud divisions have been growing faster than the rest of their business for years. When cloud revenue is mentioned there, a significant part of it is embedded in such platform services.
