Synopsys and Ansys power the future of innovation—connecting silicon to systems.
EDA
System
IP
SolvNetPlus
Training & Education
Part 1: Hitting the Capacity Wall
Part 2: The Rise of Autonomous Engineering
For much of the semiconductor industry's history, the foundations of competitive advantage were straightforward: larger teams, better tools, more compute, and more efficient processes.
As agentic AI systems take on a greater share of the work, two new differentiators are emerging. The first is the expertise those systems can draw upon. The second is how effectively that knowledge can be applied across the full chip design flow.
Delivering that combination is difficult. It requires deep domain expertise embedded in the platform as well as usable knowledge from the organizations that deploy it.
Enhance Your Chip Design with AI
Explore the enhanced Synopsys.ai brochure, featuring cutting-edge advancements in Advanced Optimization, Generative AI, and Agentic AI to transform your chip design process.
Not all agentic AI systems are equal
Complex engineering problems are rarely solved through generic or repurposed information. They require specialized reasoning, domain-specific context, and fluency in the requirements, constraints, and tradeoffs that shape engineering decisions.
A general-purpose agent connected to isolated tools is very different from an agentic system integrated across the full design flow, spanning frontend, backend, and analog design as well as verification, signoff, manufacturing, and multiphysics simulation and analysis. Full-flow systems need to understand how choices made in one stage can create downstream effects for timing closure, power, area, reliability, and manufacturability. Although mainstream agentic AI systems are becoming widely available, they don’t offer the domain specificity, deep expertise, and full-flow integration required for complex chip design.
Agentic AI systems must also be able to leverage the company-specific knowledge that makes each engineering organization distinct: its methodologies, design history, priorities, best practices, and accumulated experience.
The strongest results come when these two forms of knowledge work together: embedded domain expertise to understand chip design, and company-specific expertise to act effectively within a particular engineering environment.
This becomes especially important as systems move from recommending actions to executing workflows. Agentic systems need to reflect sound engineering judgment in the actions they take, including which paths are worth pursuing, which tradeoffs are acceptable, and which outcomes meet the organization’s design and project requirements.
The most valuable knowledge is often the hardest to find
The problem is that much of the semiconductor industry's most important expertise was never designed to be consumed by a machine.
Some knowledge exists in documentation, specifications, test results, and design databases. But a significant portion resides in the minds of human engineers — the instincts developed through years of debugging difficult failures, the shortcuts used to navigate complexity, and the judgment required to recognize patterns, identify root causes, and choose the most appropriate path forward when several options appear equally valid.
Consider how seasoned engineers approach a congestion issue, a verification failure, or a stubborn implementation bottleneck. Experts know where to look first, which signals matter most, and what sequence of actions is most likely to resolve the problem efficiently.
These decisions — and the context and rationale behind them — are rarely captured in manuals or process documentation.
Many organizations possess extraordinary expertise, but they struggle to apply it consistently. Knowledge remains fragmented across teams, embedded in individuals, or accumulated through experience without ever becoming part of a broader system.
As autonomous engineering expands, these limitations become more consequential. An AI system can only act on the knowledge it can access.
From stored knowledge to operationalized expertise
The challenge is making human expertise usable by autonomous systems. Organizations can do so in four practical ways:
- Capture: Record the rationale behind expert decisions in addition to outcomes. This includes the signals engineers trusted, the alternatives they rejected, the constraints they prioritized, and the tradeoffs that shaped the decision.
- Structure: Organize knowledge around the design stages, failure patterns, tool outputs, metrics, fixes, decision points, and outcomes where it is most relevant.
- Connect: Embed knowledge directly into engineering workflows so it appears when an agentic system needs to diagnose, decide, act, or iterate.
- Govern and refine: Validate useful knowledge, update changing practices, retire outdated guidance, and incorporate expert feedback as tools, design rules, and architectures evolve.
None of this is easy. It requires better documentation, new practices for capturing engineering rationale, processes for maintaining knowledge over time, and a culture that treats expertise as a shared asset.
Despite the investment and change required, those efforts make expertise usable by autonomous systems at the point of decision. Applied consistently across the full design flow, that knowledge becomes a durable source of competitive advantage.
Part 1: Hitting the Capacity Wall
Part 2: The Rise of Autonomous Engineering
Part 4: From Execution to Orchestration (coming soon)
Continue Reading
Agentic AI in Chip Design: The Rise of Autonomous Engineering (Part 2 of 4)
