AI through Visuals - Hardware

Most AI content today feels flat.

We read threads.
We skim blog posts.
We copy prompts.

But very rarely do we actually see how AI systems work.

Why I’m starting this series

As engineers, our edge is not just using tools.

It’s understanding systems.

Lately, I’ve been thinking a lot about this:

Are we still thinking deeply about what we’re building…
or just orchestrating tools we don’t fully understand?

That’s where this idea came from.

Introducing: AI Visual Series

I’m starting a series of interactive visual explainers to break down AI concepts:

• infrastructure
• systems
• tradeoffs
• bottlenecks
• real-world constraints

Not with walls of text…

But with visuals you can explore and interact with.

👉 You can explore the series here

The goal isn’t about simplifying AI.

It’s about making complex systems intuitive through visuals.

Because once you see something you think differently about it.

Without further delay, let's explore the first post in this series.

AI Hardware, Explained Visually

We talk about AI like it’s software.

Prompts. Models. APIs.

But modern AI doesn’t run on “code”.

It runs on a massive physical stack of hardware.

1. The AI hardware stack

A modern AI system isn’t just a server.

It’s a layered system:

• GPU / accelerator
• CPU
• high-bandwidth memory (HBM)
• advanced packaging
• networking
• storage
• power
• cooling

Each layer matters.

Each layer can break.

2. A global system

What surprised me most:

This stack is not built in one place.

• Design → United States
• Fabrication → Taiwan
• Memory → South Korea
• Lithography → Netherlands
• Materials → Japan
• Assembly + deployment → China + Southeast Asia

No single country controls the full system.

3. Where things actually break

We used to think scaling AI meant just adding more GPUs

But that’s no longer true.

The real bottlenecks are now:

• HBM memory
• advanced packaging
• networking
• power availability
• cooling

And here’s the counterintuitive part:

Compute itself is no longer the main constraint.

The key takeaway

The more I dig into AI…

The less it feels like software engineering.

And the more it feels like:

• distributed systems
• hardware engineering
• energy infrastructure

All at once.

Why this matters for us

If you’re building with AI today:

• you’re sitting on top of this entire stack
• you’re affected by its constraints
• you’re making decisions that depend on it

Understanding it, even at a high level, changes how you think.

You can play around with the interactive visuals here

What’s next in the series

I’ll keep building these visual explainers around:

• evolution of AI chips
• cost of a prompt
• model capability vs compute
• open vs closed AI ecosystems

Curious to hear:

👉 what AI concepts would you like to see visualized next?