en.Wedoany.com Reported - Cadence has recently launched a series of engineering tools driven by Agentic AI, with the core philosophy not to replace engineers, but to teach AI how to use the company's four decades of accumulated Electronic Design Automation (EDA) tools and expertise.

Rob Knoth, Group Director of Product Management for System & Verification at Cadence, discussed this development in a recent exchange. He emphasized that the goal is not to make AI a comprehensive chip design expert, but to enable it to call upon existing, mature EDA tools such as SPICE simulators, formal verification engines, and timing signoff tools. These tools themselves contain years of algorithmic and engineering knowledge.

Cadence's development in AI-driven EDA has progressed through several stages. Even before the generative AI boom, Cadence had embedded machine learning and reinforcement learning in multiple tools to optimize metrics such as power, performance, and area. Subsequently, Cadence added natural language interfaces to individual tools, allowing engineers to interact with them through conversation. The latest stage introduces Agentic AI super-agents to orchestrate these tools.

Cadence divides the engineering domain into different "stacks." ChipStack focuses on digital design creation and verification, ViraStack targets custom analog design, InnoStack handles digital implementation and signoff, while 3DStack and SystemStack extend the same concept to advanced packaging and complete system design. At the top of each stack sits an Agentic AI super-agent that coordinates multiple specialized sub-agents, which are responsible for invoking the corresponding Cadence tools and determining the next steps.

This approach aims to amplify engineers' capabilities. For example, engineers can use AI to run thousands of SPICE simulations instead of hundreds, thereby exploring more design alternatives. Rob Knoth noted that AI does not exclude engineers from the loop but rather provides them with more powerful tools.

Looking ahead, Rob Knoth stated that the current super-agents primarily handle orchestration and coordination tasks. In the future, the system may become more autonomous, with the ability to self-improve and create entirely new sub-agents. Cadence's strategic core is to "teach AI how to use forty years of EDA," rather than attempting to replace it with AI.