You don’t need a monster GPU to run AI locally. A tiny microcontroller can already handle basic models, while a Raspberry Pi, laptop, home server, or GPU can take you all the way to speech, vision, coding, image generation, and video.

The key is understanding just two things: how much memory you have, and how fast your hardware can move that memory.

We’ll talk about:

• From ESP32 to Raspberry Pi: What tiny devices can actually run, and where microcontrollers hit their limits.
• Why 8GB Isn’t Always 8GB: How the Raspberry Pi 5 and iPhone 16 can have the same RAM but very different AI performance.
• The RAM Formula: A simple way to estimate how large an LLM your laptop or desktop can fit.
• 32GB, 64GB, 128GB:  ...