Workshop

AI PCB Design: What I'm Missing

Published on: 2026-02-26

By: Ian McCutcheon

AI PCB Design: What I'm Missing

I don't claim expertise in AI-driven PCB design tools. What I do have is experience with the frustrating reality of modern PCB layout workflows — and a clear sense of what's missing.

Here's the pattern I keep hitting:

You start with a schematic. That's good. The tool guides you through component placement, pin assignments, the logical connections. Everything makes sense on paper.

Then you move to the board layout.

You place your components. You start routing traces. And somewhere in that process, you hit a wall. A trace from point A to point B becomes impossible given your layer count, your component density, your constraints. Often it's a pin on a dense MCU — an ESP32, or something even more cramped — where GPIO 12 simply cannot reach where it needs to go.

But: GPIO 12 and GPIO 32 are functionally equivalent. If you swapped them in your schematic, the routing problem might vanish entirely. Or if you flipped the chip to the other side of the board, the entire routing topology could change.

The tools don't tell you this.


What I'm Looking For

What I want from an AI PCB routing tool isn't just "here's a route" or "this route is impossible."

I want feedback that suggests alternatives:

That level of intelligence — understanding functional equivalence, suggesting schematic-level changes that simplify layout — is where AI PCB tools need to go to become genuinely useful.

Right now, most tools treat schematic and layout as separate domains. The AI sees a routing problem and says "can't do it" or "needs more layers." It doesn't look back at the schematic and say "change this, and the problem disappears."


Is This Out There?

I'm asking because I genuinely don't know.

If there's a tool that does this — that provides schematic-aware routing feedback — I'd love to hear about it. If you're building one, or if you've seen this capability somewhere, please drop a note in the GitHub repo discussions or open an issue.

This isn't just about convenience. It's about closing the loop between design intent and physical reality. The schematic represents what you want to build. The layout represents what's actually possible. Right now, that feedback is manual, slow, and often discovered too late.

AI should be able to bridge that gap.

I'm having déjà vu, so if anybody's heard this same query from me before, please tell me when I said it out loud before.