A new development in semiconductor design suggests that artificial intelligence is moving beyond assistance and into full autonomy. The IEEE Spectrum article examines a system built by startup Verkor.io that uses “agentic AI” to design an entire CPU core from scratch, marking a significant milestone in chip engineering.
The system, called Design Conductor, orchestrates large language models to carry out each stage of chip development. Starting from a short text prompt, it progresses through architectural planning, coding at the register-transfer level, debugging, and layout generation. This mirrors the workflow traditionally handled by teams of engineers but compresses it into a tightly coordinated automated process.
The resulting processor, named VerCore, is based on the open-standard RISC-V architecture. It operates at roughly 1.48 gigahertz and achieves performance comparable to a laptop CPU from around 2011. While modest by modern standards, it is the first example of an AI system delivering a complete CPU design rather than isolated components.
Speed is one of the most striking aspects. The AI system produced the full design in about 12 hours, demonstrating how development cycles could shrink dramatically. However, the chip currently exists only in simulation, validated using standard tools such as the Spike simulator and an academic 7-nanometer design kit.
Despite the breakthrough, limitations remain. The AI lacks human intuition and can pursue inefficient design paths, sometimes requiring extensive iteration to resolve issues. Engineers still play a critical role in guiding, validating, and refining results.
The broader implication is not the replacement of engineers but a shift in how chips are created. Agentic AI could enable smaller teams to design complex processors faster and at lower cost, fundamentally changing the economics and pace of semiconductor innovation.