AI & power electronics: states of the art

11 May 2025 - Carrboro

AI for power electronics

In this section, I will list Ai tools designed for power electronics.

I am not convinced yet that we really need AI based control algorithms for our power systems; this collection will mostly be for off-line designs.

DeepPCB: Website link is the AI-powered, fully autonomous printed circuit board tool that accelerates the design process and ups industrial-grade PCB performance. Create and scale your product without compromise, and only pay for what you use.
CircuitSynth. Paper link. Circuit topology synthesis using LLM.
frenetic.ai: Website link Design power supply in minutes - with AI.
AI-Mag. ETH work. Link. Inductor Modeling and Design with FEM/ANN.
PE-GPT. Repo here. I am still questioning the significance of such GPT-like models over general-purpose models such as chatgpt / gemini.
Circuit Mind. Their claim of speeding up by 75% seems to be more conservative than Deep PCB, thus more trustworthy (to me)?

What energy source do we need for training AI? Is it true the nuclear plants will be our only solution?
Is power electronics still a subject of science?
Why is everyone trying to eliminate their work needs by AI? This is a terrifying game, indeed. The first one gets the most by killing all the rest. Not just in power electronics – biology, mechanical eng, etc., even machine learning itself.