How Agentic AI Chip Design Built a Full RISC-V Core

How Agentic AI Chip Design Built a Full RISC-V Core

How Agentic AI Chip Design Built a Full RISC-V Core

https://spectrum.ieee.org/ai-chip-design

Publish Date: 2026-04-22 07:00:01

Source Domain: spectrum.ieee.org

Summary:

The article delves into the progress of AI-driven chip design since 2020, highlighting the advancements by Verkor.io’s agentic AI system known as Design Conductor. Initially, researchers fine-tuned GPT-2 for basic logic circuit fragments, moved to 2023’s GPT-4-driven designs for an 8-bit processor, then progressed to more generalized LLMs capable of basic chip trials by 2024. In 2025, Verkor debuted VerCor, a RISC-V CPU core entirely designed by Design Conductor from initial specifications autonomously. Though VerCore operates at 1.48 GHz and resembles mid-2010s processor performance, it represents the first complete RISC-V design by an AI system. While not a competitor to current market leaders, it signals AI’s expanding role in chip design, suggesting potential for faster iterations and smaller, less-resourced teams to partake in the design process. Design Conductor emulates human engineers’ workflow but falls short in intuition and efficiency for complex designs, emphasizing the continued need for human expertise.

Key Points:

  • Advanced AI models began with logic circuit fragments in 2020 and progressed to designing an 8-bit processor using GPT-4 by 2023.
  • Verkor.io’s Design Conductor autonomously designed a RISC-V CPU core, VerCore, achieving a complete design, a first for AI agents.
  • VerCore, with a clock speed of 1.48 GHz, performs similarly to CPUs from 2011 and highlights AI’s expanding capabilities rather than presenting immediate competition.
  • Design Conductor mirrors human workflow with structured steps like design, implementation, and testing, managed autonomously from an initial specification.
  • While promising in accelerating iterations and making chip design accessible to smaller teams, Design Conductor still relies heavily on human oversight and guidance for complex designs.