How computers physically work

Before we talk about AI at all, we need to talk about the machine it lives on. AI isn't floating in "the cloud" as some abstract force — it's running on real, physical hardware in a real building somewhere, made of parts you can hold in your hand. Once you can picture those parts, AI gets a lot less mysterious.

Every computer, from your laptop to the giant servers running Agentforce, is built from the same four pieces.

The four parts

The CPU — the worker. The CPU (Central Processing Unit) is the part that actually does things. Think of it as a single, very fast worker following instructions one after another: add these numbers, compare these two values, move this here. It's brilliant at doing complex tasks in sequence, but it's essentially one worker (modern CPUs have a handful of "cores," so really a small team).

RAM — the desk. RAM (Random Access Memory) is the workspace. Picture a desk where the worker spreads out everything they're using right now. It's extremely fast to reach, but small and temporary — clear the desk (turn off the power) and everything on it is gone. The more RAM you have, the more you can work on at once.

Storage — the filing cabinet. Your hard drive or SSD is long-term storage. It's much bigger than the desk and it keeps things when the power is off — but it's slower to reach. So the worker pulls what they need out of the filing cabinet (storage), puts it on the desk (RAM), works on it, and files the results back.

The GPU — the room full of workers. The GPU (Graphics Processing Unit) started life drawing video-game graphics, which requires doing the same simple math on millions of pixels at once. So instead of one fast worker, a GPU is thousands of simpler workers all doing math in parallel. Hold that thought — it's the single most important idea for understanding AI, and the next page is all about it.

How data actually flows

Here's the loop that every program — including AI — runs on:

1. Data sits in storage (the filing cabinet).
2. The computer loads what it needs into RAM (onto the desk).
3. The CPU (or GPU) reads that data, processes it, and writes results back to RAM.
4. Results get saved back to storage, or sent to your screen.

That's it. Move data into the fast workspace, process it, move the results back out. If you've ever heard a programmer talk about "loading something into memory," this is what they mean — and underneath it all, the CPU is executing millions of tiny instructions a second to make it happen.