What is EDA Agentic AI?

EDA Agentic AI systems are built upon a framework of autonomous agents, each will be programmed to perform distinct engineering tasks and adapt dynamically to changing requirements. Unlike rule-based automation, these agents will possess capabilities for reasoning, planning, learning, and execution. The workflow typically involves several key steps and features:

Reasoning and Planning

Agents begin by analyzing the design goals, constraints, and available resources. They develop intelligent, adaptable plans of action, taking into account evolving project requirements. For example, an agent might evaluate trade-offs between performance, power, and area for a given chip design, then propose optimal strategies for RTL generation or verification planning.

Learning and Improvement

A hallmark of Agentic AI is its ability to learn from experience. After each project or task, agents receive feedback on their performance and outcomes. This continuous learning loop enables them to refine their strategies, improve efficiency, and adapt to new challenges. Over time, Agentic AI systems become increasingly effective, reducing errors and accelerating future projects.

Execution and Orchestration

Agents autonomously will carry out engineering tasks, such as generating RTL code, creating verification testbenches, planning simulation runs, and managing data flows. They will work individually or in coordinated teams, orchestrating complex workflows that span multiple stages of the design lifecycle. Human engineers will maintain oversight and can intervene at critical decision points, but much of the routine, computationally intensive work will be handled by the agents.

Multi-Agent Collaboration

One of the most powerful aspects of EDA Agentic AI is the ability for multiple agents to collaborate. In a multi-agent system, agents will communicate, share data, and synchronize their activities to tackle large-scale engineering challenges. For example, one agent may focus on logic synthesis while another handles verification planning, both working in tandem to ensure design integrity and optimal outcomes.

Human-AI Partnership

While agents will take on repetitive and complex tasks, human engineers will be freed to focus on higher-level activities—such as strategic planning, innovation, and problem-solving. This partnership between humans and AI agents leads to better decision-making, faster project delivery, and improved overall product quality.

Levels of Agentic AI

Agentic AI will become increasingly more capable and can be categories into 5 levels, L1 to L5, with each level characterized by increasing capabilities of AI agents. The figure below shows the progression from basic, assistive automation (L1) to collaborative, partially autonomous agents (L2–L4), and ultimately to highly autonomous, self-directed agents (L5), where human engineers rely on AI agents for a myriad of engineering tasks.