Siemens and NVIDIA Build Self-Verifying AI Agents for Chip Design
New collaboration embeds physics-based validation into autonomous workflows for semiconductor and PCB engineering teams.
Siemens has expanded its partnership with NVIDIA to bring self-verifying artificial intelligence agents to electronic design automation, targeting the complex workflows that semiconductor and printed circuit board engineering teams navigate daily.
The enhanced collaboration centers on Siemens' Fuse EDA AI Agent system, which now integrates NVIDIA's AI infrastructure to enable autonomous agents that can reason, act, and continuously validate their decisions against deterministic, physics-based engineering tools. The approach addresses a fundamental challenge in applying AI to chip design: ensuring that autonomous systems produce results engineers can trust.
How the technology works
The updated Fuse system combines several NVIDIA technologies with Siemens' domain expertise. Agents built using NVIDIA's NeMo Gym open library optimize for result quality, speed, and token efficiency—learning from each project to refine execution strategies over time. NVIDIA's OpenShell secure runtime provides enterprise-grade security and access controls for agents operating across entire EDA environments.
The system employs NVIDIA's Nemotron models for advanced reasoning, enabling agents to evaluate complex engineering trade-offs while maintaining token efficiency. NVIDIA's accelerated computing and CUDA-X libraries power both AI reasoning and EDA engines, compressing timelines from days to hours without sacrificing accuracy.
These capabilities support multi-agent coordination across the full semiconductor lifecycle, from high-level synthesis through physical implementation and signoff verification. The agents work with Siemens' existing EDA portfolio, including Catapult for synthesis, Questa One for verification, Solido for custom IC design, and Calibre for signoff.
Custom IC design gains
Siemens highlighted specific improvements in custom integrated circuit workflows. The company's Solido Characterization Suite now automates library characterization with agentic AI, generating and verifying Liberty files while reducing characterization turnaround times by more than 10X and cutting token costs by 5X to 10X.
A new Solido Layout Analyzer tool adds AI-powered analysis of parasitic and layout-dependent effects in post-layout designs, accepting natural language prompts for results analysis and fix recommendations. STMicroelectronics' non-volatile memory team noted the tool brings layout insight earlier into the design flow, potentially reducing debugging time by weeks.
Verification acceleration
In digital verification—which consumes up to 70 percent of design time—Siemens is deploying NVIDIA's Nemotron 3 Ultra reasoning model through its Questa One Agentic Toolkit. The model's performance in agentic RTL benchmarking enables autonomous agents to evaluate engineering trade-offs while validating designs against reference test harnesses.
"We're at an inflection point where the complexity of AI chips, chiplets, and 3D ICs has outpaced traditional verification methodologies," said Abhi Kolpekwar, senior vice president at Siemens EDA. The agentic approach aims to orchestrate multi-domain verification and reason across billions of test scenarios.
The system integrates with Siemens' Intelligence Center X, the company's enterprise industrial AI environment that coordinates processes across design, manufacturing, and supply chain operations.
Why it matters
Semiconductor design complexity has reached a point where traditional automation struggles to keep pace. Self-verifying AI agents represent an architectural shift: rather than simply automating individual tasks, these systems can orchestrate entire workflows while continuously checking their work against proven engineering tools. For an industry where a single design error can cost millions and delay product launches by months, the ability to accelerate development while maintaining—or improving—quality and trust could reshape competitive dynamics. The integration with broader enterprise systems also signals how AI in chip design will increasingly connect to manufacturing and supply chain intelligence.
The enhanced capabilities will be available in forthcoming releases of Siemens' AI-native EDA portfolio, according to details first reported in the company's announcement.
This is an original analysis by the Omega editorial team. Source reporting: AI Watch.
Want systems like this working for your business?
Book a Call
