Printed circuit board (PCB) designs have become more complex even as the chips that populate the boards have grown with chiplets, leading to ever larger and hotter packages. It’s estimated that over 60% of an engineer’s time is not spent designing; rather it involves addressing tasks that can be automated with the help of agentic AI (Fig. 1).

This is where Cadence’s AuraStack AI Super Agent (Fig. 2) comes into play. You can watch a short demo in the video above. AuraStack is hooked into the company’s design tools, allowing it to collect and analyze architectural details to assist a developer. With this human-in-the-loop approach, the developer is able to make queries about the design and receive responses such as what types of improvements can be made based on system analysis.

The agent not only provides feedback, but it implements changes upon request. While many of the tools incorporate AI internally, they provide the deterministic design implementation. The agentic AI initiates the same type of operations that would be done by a designer. The tools handle details like routing and placement based on user prompts (Fig. 3). The video shows examples of this type of interaction.

The AuraStack AI Super Agent is just one of Cadence’s AgentStacks (Fig. 4). ChipStack addresses chip-level design and verification while ViraTack helps analog designers. The InnoStack agent focuses on synthesis and place-and-route through signoff analysis and engineering change order (ECO) execution.