server-side

When you open a website or use an app, two computers are always involved. One is your device, so your phone, tablet, or laptop. The other stands in a data center, a hall full of powerful computers, and is called a server. “Server-side” means that a certain step of the work happens on this remote computer and not with you. The counterpart is called “client-side”: there your own device does the computing. One example: the search results on Google are generated server-side, whereas scrolling through the list is handled by your phone itself.

Why the work often doesn’t stay on your device

The most important reason is computing power. A current language model, meaning an AI system that writes texts, needs specialized graphics chips costing many thousands of euros. Such hardware doesn’t fit into any phone. So the computational work runs in the data center, and your device only displays the finished text.

The second reason is control. Anything that arrives on your device can, in principle, be read out and altered there. That’s why an online shop checks server-side whether you have actually paid. If this check took place in the browser, it could be circumvented with a few clicks. Companies therefore deliberately keep important rules and secrets on their own side.

The third reason is currency. If the provider changes something on the server, the change takes effect immediately for all users. With client-side software, everyone would first have to install an update. That’s exactly why providers are increasingly shifting functions back into their own infrastructure.

The journey of a request into the data center

The process is always similar. Your device packages your input into a request and sends it off over the internet. The server accepts it, computes, perhaps queries a database as well, and sends a response back. So your device usually only gets text or image data, not the program behind it.

So that this doesn’t become chaotic, there are fixed handover points. They’re called interfaces, or also APIs: a clearly defined list of requests that a server understands. An app knows this list and adheres to it. What the server does with it internally remains invisible to the app.

One drawback is unavoidable: the journey costs time. This delay is called latency and is often around a few dozen milliseconds. With a chatbot, you barely notice it. With a fast online game, it’s crucial, which is why part of the computation is deliberately kept on the device there. Without an internet connection, server-side processing doesn’t work at all.

From the camera app to the cloud bill

You encounter this principle constantly in everyday life. The translation of a text, the speech recognition of your voice assistant, and the recommendations on Netflix are generated server-side. Also, an image filter that completely reshapes a photo is often uploaded and computed outside your phone. Smaller tasks like recognizing faces in a photo album, on the other hand, now often run directly on the device.

In business news, the term usually comes up in connection with costs. Server-side AI incurs computing costs with every single user request. That’s why media report on billions in investments in data centers and on the providers' electricity consumption. Whoever shifts functions onto users' devices shifts these costs to the customers.

A common misconception: “server-side” doesn’t automatically mean “secure.” Your data then resides with a foreign company, often in another country. For the technology, the server is more protected, but not necessarily for your privacy. It’s exactly this distinction that a large part of the debate about data protection in the cloud revolves around.