When KiCad 9.0 was released, the most common comment in hardware engineering circles was: “AI is going to route our boards for us.”
Three months later, those who’ve actually used AI‑assisted PCB design are a lot less worried.
Not because AI is weak — quite the opposite. AI has exceeded expectations in routing, symbol generation, and DFM checking. But engineers have realized something more important: the tasks AI can replace are exactly the repetitive ones nobody wanted to do anyway. The truly valuable work is still out of AI’s reach.
This article isn’t about the grand question of “will AI replace engineers?” Let’s talk specifics: what can AI actually do in PCB design in 2026? What can’t it do? And how should you use it right now?
1. What Can AI Do for You Today?
1. Symbol and Footprint Generation
The most tedious part of schematic capture used to be flipping through datasheets page by page to manually create symbols. KiCad Copilot can read a datasheet’s pinout diagram and generate a usable schematic symbol automatically.
It’s not 100% accurate — but it saves 70% of the time. The remaining 30% is manual verification, which is still much faster than starting from scratch.
2. Repetitive Routing Tasks
Auto‑routing isn’t new, but AI has made it smarter. Old auto‑routers followed fixed rules; AI‑driven routers follow intent — tell it “this is a DDR data line,” and it knows to match lengths, control impedance, and avoid clock traces.
For medium‑complexity boards, AI routing already achieves over 90% completion on simple tasks.
3. Real‑time DRC and DFM Pre‑checks
KiCad’s AI plugins can flag issues as you draw — vias too close to pads, silkscreen overlapping traces, thermal pads without mask openings. Problems get caught during layout, not just before tape‑out.
4. Alternative Component Recommendations
When a capacitor is out of stock, AI can suggest three alternatives based on package, capacitance, and voltage rating — and even flag lead times and price trends.
2. What Can’t AI Do Yet?
1. Understand “Why”
AI doesn’t know why a trace took a detour — maybe to avoid a hot spot, maybe to leave room for assembly, maybe for EMC reasons. It only knows “this follows the rules.” Anything outside the rules is invisible to AI.
2. Handle Ambiguity
“This board might go into a car or industrial equipment — design for the stricter case” is not a prompt AI can parse. It needs explicit parameters. Hardware engineers spend most of their time dealing with things that haven’t been parameterized yet.
3. Take Responsibility
When something goes wrong, the customer comes to you, not the AI. AI has no concept of accountability — its answers are always “recommendations based on current information.” The final decision must stay with you.
3. How Should You Use AI?
Treat AI as a senior assistant, not a replacement.
AI is good at execution, not judgment. Your value in the design phase isn’t how fast you route — it’s deciding which traces go on the outer layer, which power rails need wider copper, which signals need more clearance. Those judgments are still yours.
Use the time AI saves for things AI can’t do.
If AI cuts routing time from three days to half a day, spend the extra 2.5 days on thermal simulation, signal integrity analysis, and DFM reviews — your board’s success rate will improve dramatically.
Learn to review, not just accept.
AI‑generated symbols need checking. AI‑routed traces need adjusting. AI‑recommended parts need verifying. The more powerful the tool, the more important the review skill becomes.
The Bottom Line
AI won’t make hardware engineers obsolete. But engineers who use AI will definitely make life harder for those who don’t.
This isn’t about replacement. It’s about tool evolution.
Just like the transition from manual routing to CAD tools, CAD didn’t eliminate hardware engineers — it eliminated those who refused to learn CAD.
AnyPCBA, founded in 2011, focuses on small‑to‑medium batch PCB manufacturing and PCBA assembly. We see AI‑assisted designs every day. Whether your design was drawn by hand or generated by AI, we run a full DFM review before production.
👉 AnyPCBA website: https://www.anypcba.com/
Small‑to‑medium batch PCB & PCBA | 5–5,000 pieces | Prototype to Production
Top comments (0)