Silicon is starting to design silicon — how AI is being used in chipmaking, from EDA tools to OpenAI's Jalapeño and beyond

In August, Architect Labs announced a milestone in semiconductor design: a chip almost entirely created by artificial intelligence. AI is increasingly used throughout the chipmaking process, from optimizing floorplans to generating RTL code. Advanced agentic systems can even operate EDA tools, analyze results, and make design modifications autonomously.

However, human engineers still define the chip's architecture and major decisions. This creates a feedback loop where AI models, built on processors designed by humans, help design more capable AI systems.

EDA vendors and semiconductor companies have been incorporating AI into various stages of design. Early AI-enhanced EDA tools utilized machine learning and reinforcement learning to optimize implementation options while human-defined design constraints guided the process. More recent tools can write or modify RTL and verification code, analyze reports, identify failures, and suggest fixes.

Emerging agentic systems can operate multiple EDA tools with minimal human intervention, automating parts of the chip development workflow.

While Google, Nvidia, and OpenAI have all utilized AI in their chip development processes, their levels of AI involvement differ. Google's AlphaChip uses reinforcement learning for physical floorplanning but does not design entire architectures. Nvidia trains specialized models on its proprietary engineering data, automating tasks typically performed by hardware engineers without creating autonomous chip designs.

OpenAI's Jalapeño stands out as a real-world example, directly involving AI in implementation, design-space exploration, verification, and optimization, resulting in significant performance improvements and reduced area. These advancements showcase AI's growing role in chip design, from assisting human engineers to potentially automating the entire process in the future.