
IaaS
IaaS stands for "Infrastructure as a Service" and means: A company rents computers, storage space, and network capacity in someone else's data center instead of buying its own computers. Payment is usually based on usage time, similar to electricity.
Anyone who wants to run a website, an app, or an AI program needs computers that run around the clock. In the past, you had to buy these computers yourself, set them up in a room, cool them, and maintain them. IaaS is the alternative: you rent computing power and storage space from a large provider that owns huge halls full of computers. The English name is “Infrastructure as a Service,” meaning roughly “infrastructure as a service.” Payment is based on consumption, often by the hour or even by the minute. You only rent the bare technology – whatever runs on top of it, you set up yourself.
Why companies no longer build their own server rooms
Buying computers for a company costs a lot of money at once. You also have to guess in advance how much capacity you’ll need in two years. If you guess too low, the website collapses at the first big rush. If you guess too high, expensive machines sit around unused and lose value.
Rented infrastructure solves this problem because it can be adjusted quickly. An online shop can book twenty times as many computers for a single day on Black Friday and then give them up again afterward. Experts call this flexibility scalability. For young companies, this is often crucial: they start with just a few euros a month and grow without new hardware.
What matters to investors is the flip side. The market is dominated by a few very large providers, above all Amazon, Microsoft, and Google. This business is extremely profitable and finances a large part of current AI investments. At the same time, a dependency arises, because switching to another provider is cumbersome and expensive.
What you’re actually renting from the provider
Technically, the data centers house normal but very powerful hardware. Software runs on it that divides a physical computer into many separate portions. Each of these portions behaves like its own computer with its own operating system. It’s called a virtual machine, and that’s exactly the kind of portion the customer rents.
Orders aren’t placed by phone but via a website or a short script. Within seconds, you get a new virtual machine along with access credentials. You also choose storage space for files and databases as well as network settings, such as which requests are allowed from outside. For AI applications, you additionally rent graphics cards, because these handle the types of calculations for neural networks especially quickly.
The division of labor is clearly regulated. The provider takes care of the building, power, cooling, and defective components. The customer is responsible for everything they install themselves, meaning operating system updates, programs, and security settings. This is the difference from Platform as a Service, where the provider also maintains the software environment, and from Software as a Service, where you simply use a finished program in the browser.
IaaS in the news and in everyday life
The best-known offerings are called Amazon Web Services, Microsoft Azure, and Google Cloud. In these corporations' quarterly reports, the cloud business is often the fastest-growing segment. When the news talks about billions for new data centers, it’s usually about exactly this infrastructure. European providers like OVHcloud or Deutsche Telekom are also in the game, especially when data is meant to stay in Europe.
You also encounter IaaS indirectly every day. Streaming services, banking apps, and online games almost all run on rented computers. If a large data center goes down, hundreds of websites become unreachable at the same time. Such outages regularly make headlines and show how strongly the internet depends on a few providers.
A common misconception is that rented infrastructure is automatically cheaper. With steady, permanently high load, owning your own computers can be less expensive, and some companies pull back for this reason. It gets particularly expensive when large amounts of data are pulled back out of the cloud. Anyone comparing prices should factor in such fees.