JustUpdateOnline.com – As the semiconductor industry grapples with increasingly complex system-on-chip (SoC) designs, the traditional methods of improving electronic design automation (EDA) are reaching their limits. For decades, progress was measured by the raw speed of simulators and the capacity of formal tools. However, the modern bottleneck has shifted from computational power to the intricate coordination of workflows, prompting a move toward “agentic AI” to streamline the verification process.
The Talent Crisis in the Global Chip Market
The need for smarter workflows is most evident in Asia, where a massive surge in semiconductor investment is colliding with a significant shortage of specialized labor. In China, the push for domestic self-sufficiency in integrated circuit (IC) design has outpaced the available workforce, with some estimates placing the talent deficit in the hundreds of thousands.
This challenge is also a primary focus of government policy across Southeast Asia. Malaysia has launched a National Semiconductor Strategy aimed at upskilling 60,000 high-level engineers to move into advanced packaging and equipment manufacturing. Similarly, Vietnam is working toward a goal of 50,000 semiconductor professionals by 2030. Because training new experts takes years, the industry is looking for ways to amplify the productivity of existing teams without sacrificing the rigor required for final product approval.
From Static Automation to Workflow Intelligence
Standard automation typically relies on fixed inputs and predictable sequences. However, RTL (Register Transfer Level) verification is a fluid environment where specifications often shift and new risks are discovered mid-process. This is where agentic AI offers a distinct advantage.
Unlike older tools that focus on isolated tasks, agentic systems operate across the entire verification lifecycle. These AI agents can monitor the current state of a design, formulate a plan of action, execute necessary tasks, and provide concise summaries of the results. By integrating these agents directly into the existing toolchain—rather than having them sit as external script generators—engineering teams can ensure that AI-driven actions are validated against the same strict coverage models and semantic checks used for human-led sign-offs.
Maintaining the Human Element
Despite the advanced capabilities of agentic systems, industry experts emphasize that human oversight remains a non-negotiable component of chip verification. The objective is not to replace the engineer but to remove the burden of manual coordination.
Verification often involves navigating ambiguous specifications and making subjective trade-offs that require years of experience. Determining when a design is “safe enough” for production is rarely a simple binary choice. Consequently, effective AI implementations are designed to be human-centric. The agents handle the heavy lifting of data analysis and task execution, but the final authority on intent, scope, and sign-off remains firmly in the hands of the engineering team.
Building a Foundation of Trust
For agentic AI to be effective, the underlying architecture of EDA tools must evolve. This requires engine-native interfaces that allow AI agents to interact directly with simulation data and system states in a structured manner. By maintaining a continuous context of historical results, test benches, and design iterations, these systems can reason across multiple cycles of the development process.
To mitigate the risks of AI—such as the potential for error propagation—developers are implementing "bounded actions." This framework restricts the AI to specific, validated tasks and requires human approval at critical decision points.
The Future of Semiconductor Design
As chip complexity continues to escalate, the industry is shifting its focus from how fast a tool can run to how intelligently a workflow can be managed. Agentic AI represents the next frontier in this evolution, offering a way to manage the mounting pressure of design iterations. While the technology is still maturing, the move toward workflow intelligence is expected to provide the productivity boost necessary for verification teams to keep pace with the next generation of technological innovation.
