NovaSilicon Explores Autonomous Chip Design with AI Agents

新智元 · wechat · 2026-08-25

Startup NovaSilicon has raised a new funding round led by SenseTime's Guoxiang Capital. The company is building a superintelligence system for the chip industry, using Recursive Self-Improvement (RSI) methodology to enable AI agents to autonomously handle design, optimization, and verification tasks.

Traditional EDA tools rely on human-defined rules and search spaces, facing bottlenecks in complex chip design. NovaSilicon proposes building an "AI organization" where agents with different functions—architecture, analog design, verification, process—collaborate, shifting from "static constraints" to "dynamic cognition." Large Language Models introduce physical semantic understanding, optimizing designs based on engineering significance rather than pure mathematics.

The company has published a paper on "Thinking Boundary Aware Super Intelligence" and secured tens of millions in orders, aiming to shift the chip design cycle from a human-speed baseline to an AI-iteration-speed baseline.

Original post →

More from Companies & People

Companies & People channel →