By TradingView
Publication Date: 2026-10-14 18:10:00
AI can now help write code, summarize documents and handle increasingly complex office tasks.
Designing a semiconductor is proving much harder.
Nvidia NVDA and chip-design automation startup ChipAgents have launched a new benchmark to test how well AI models can handle the entire chip-design process, according to The Information.
The result was a reality check.
The best-performing AI agent, powered by Anthropic’s Fable 5.1, scored just 16.6% on the evaluation, according to ChipAgents.
That does not mean AI is useless for semiconductor design.
It means the technology still has a long way to go before it can reliably handle one of the most complex engineering workflows in the world with limited human supervision.
The benchmark is harder than existing chip tests
Most existing evaluations focus on narrower tasks.
ChipAgents CEO William Wang said the new benchmark is different because it spans the full design process rather than testing isolated pieces of work.
That matters because chip design is not one problem.
It is a chain of interdependent decisions covering architecture, verification, implementation and error checking. A model can perform well on one step and still fail when those steps have to work together.
ChipAgents was founded in 2024 and develops AI-agent software designed to speed up and automate chip development.
The company has raised $134 million from investors including B Capital, Micron and MediaTek.
Wang declined to identify ChipAgents’ customers or say whether Nvidia is…
