Siemens Deploys AI Agents for Chip Design
Siemens is integrating "agentic AI" into its Questa One chip design platform to speed up development cycles. The company announced the new AI workflows will help accelerate the verification process for complex integrated circuits. The move aims to help semiconductor firms keep pace with soaring demand from the AI, auto, and IoT markets.
The new "Questa One Agentic Toolkit" embeds autonomous AI agents directly into the chip verification workflow, a process that can consume up to 70% of the total chip development cycle. These agents are designed to reason, plan, and execute complex verification tasks with configurable human oversight at critical decision points. The toolkit launches with five distinct agents, including one that generates synthesizable code from natural language, another that creates entire verification plans from design specifications, and a debug agent that performs root cause analysis to accelerate error tracing. Early adopter MediaTek reported that its engineers became proficient within hours and completed tasks that typically require days. This initiative is part of Siemens' broader strategy around its Fuse EDA AI framework. The system is designed to be "framework-agnostic," allowing it to integrate with other AI platforms like GitHub Copilot and Claude Code, thereby protecting customers' existing investments in other tools. Underpinning the agentic workflows are NVIDIA's Llama Nemotron and NIM technologies. According to NVIDIA's Tim Costa, these reasoning models provide a powerful foundation to accelerate the development of advanced electronic systems. The move addresses a critical bottleneck in the semiconductor industry. As chips become more complex with 3D ICs and chiplet architectures, the "verification productivity gap" continues to widen, making traditional verification methods too slow to keep pace with design complexity. Siemens' competitors are also heavily invested in AI. Cadence, for example, has been integrating what it calls "optimization AI" and is now also moving toward agentic AI, predicting that up to 90% of advanced chips will integrate agentic AI by 2027. The toolkit is currently available through an early access program. This allows customers to validate the governance controls and integrate the AI-driven workflows into their established methodologies before a broader rollout.
