Comment by tiberius_p

I work in hardware design and verification. I've seen many AI-based EDA tools proposed at conferences but in the team that I'm working now I haven't seen AI being adopted at all. Among the proposed tools that caught my attention: generating SystemVerilog assertions from natural language prompts, generating code fixes from lint errors, generating requirements, vplans and verification metrics from specifications written in natural language, using LLMs inside IDE's as coding agents and chat bots to query the code. I think the hardware industry will be harder to penetrate by AI because hardware companies are more secretive about their HDL code and they go to great lengths to avoid leaks. That's why most of them have an in-house IT infrastructure and they avoid the cloud as much as possible especially when it comes to storing HDL code, running HDL simulations, formal verification tools and synthesis. Even if they were to employ locally hosted AI solutions that would require big investments in expensive GPUs and expensive electricity bills: the industry giants will afford it while the little players won't. The ultimate goal is to tapeout bug-free chips and AI can be a great source of bugs if not properly supervised. So humans are still the main cogs in the machine here. LLMs and coding agents can make our jobs a whole lot easier and pleasant by taking care of the boring tasks and leaving us with the higher level decisions, but they won't replace us any time soon.