AI Agent Designs a RISC-V CPU Core From Scratch

2026-04-22 · Source: IEEE Spectrum · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Advanced, short

Summary

Verkor.io, an AI chip design startup, claims to have developed the first RISC-V CPU core, dubbed VerCore, designed entirely by an agentic AI system. The VerCore CPU operates at 1.48 GHz, achieving a CoreMark score of 3,261, which positions its performance similarly to Intel's Celeron SU2300 from 2011. The AI system, named Design Conductor, is a harness for large language models (LLMs) that autonomously follows human-like chip design workflows, from an initial 219-word specification to generating a GDSII file. This system completed the VerCore design in 12 hours. While the chip has not been physically produced, it was verified in simulation using Spike and laid out with the open-source ASAP7 PDK, capable of running uCLinux. Verkor.io plans to release design files and demonstrate an FPGA implementation at DAC 2026.

Key takeaway

For research scientists exploring advanced chip design automation, you should investigate agentic AI systems like Verkor.io's Design Conductor. This approach demonstrates that AI can autonomously complete complex tasks, potentially reducing design cycles and making sophisticated chip development accessible to teams with fewer resources. Consider experimenting with structured AI agent harnesses to tackle end-to-end engineering challenges, even if human oversight remains crucial for avoiding "rabbit holes" and providing expert guidance.

Key insights

Agentic AI systems can autonomously design complete CPU cores, mirroring human engineering workflows.

Principles

• AI agents can solve entire design problems.
• Human workflows guide agentic AI systems.
• AI model capabilities improve rapidly.

Method

The Design Conductor system uses an LLM harness to follow structured steps (design, implementation, testing), managing sub-agents and a file database, to generate a GDSII file from a design specification.

In practice

• Automate chip design from spec to GDSII.
• Accelerate design iteration cycles.
• Enable smaller teams to undertake complex designs.

Topics

• AI Agent Design
• RISC-V CPU Core
• Design Conductor System
• Electronic Design Automation
• Chip Design Automation

Code references

• riscv-software-src/riscv-isa-sim
• The-OpenROAD-Project/asap7

Best for: Research Scientist, AI Scientist, AI Hardware Engineer, AI Architect

Related on AIssential

• First Take on CadenceLive and Its AI Agent Stacks for EDA — Cadence's agentic AI EDA stack uses a hierarchical orchestrator and super agents to accelerate system design.
• Exclusive: Allen Wu on Disrupting $100M Cost of Building Custom AI Chips — Heterogeneous custom AI chip development can be made affordable and accessible using an open, modular RISC-V baseline.
• We Need to Talk About AI — AI is rapidly transforming scientific research, raising profound questions about human intellectual roles and the future of discovery.
• Claude Agents Just Built a Fully Functioning C Compiler With Zero Human Input — AI agents can autonomously build complex software systems through coordinated, iterative development.
• Your AI Agent Isn’t Hallucinating. It’s Failing By Design. — Agentic AI failures are primarily engineering design flaws, not LLM hallucinations, and are largely preventable.
• Powering agentic AI sales strategy with Amazon Bedrock AgentCore — Orchestrating specialized AI agents with a unified interface significantly reduces cognitive load and improves operational efficiency.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by IEEE Spectrum.