Siemens Digital Industries Software announced the Fuse EDA AI Agent, a domain-scoped autonomous AI agent designed to plan and orchestrate multi-tool and multi-agent workflows across semiconductor, 3D IC, and printed circuit board (PCB) system design processes from conception through manufacturing sign-off. The Fuse EDA AI Agent builds on the existing Fuse EDA AI system, which includes a retrieval-augmented generation (RAG) framework, multimodal EDA data support, a secure agentic orchestration layer with Model Context Protocol (MCP), Agent Skills functionality, and an open framework for third-party integration. It supports the NVIDIA Agent Toolkit, Nemotron models, and NVIDIA AI infrastructure to enhance tool calling and reasoning.
The agent manages end-to-end automation across Siemens' EDA portfolio, covering front-end design, verification, physical implementation, and sign-off. Specific integrations include:
- Architectural exploration, design planning, and RTL coding with Catapult software.
- Testbench generation and debugging with the Questa One Agentic Toolkit.
- Place-and-route, timing closure, and power optimization with Aprisa software.
- Custom design and verification with Solido software.
- Hardware-assisted verification with the Veloce system.
- Physical verification sign-off, including DRC violation analysis, with Calibre software.
- Power/ground load optimization and signal path plan clustering in Innovator3D IC software.
- Layout, signal integrity, and analyses in Xpedition and Hyperlynx software.
- Design-for-test workflows with Tessent software.
- Optical proximity correction with Calibre OPC products.
The Fuse EDA AI Agent addresses limitations of generic AI tools in EDA environments, such as lack of domain knowledge, IP exposure risks, and context saturation. It incorporates domain-specific expertise with built-in validation and guardrails, Agent Skills for multi-step task execution via customizable playbooks, scalable AgentOps and MCP architecture for dynamic tool discovery and hierarchical planning with supervisor and worker agents, enterprise-grade resource allocation across HPC clusters, and embedded security features including role-based access controls, audit trails, and human checkpoints.
Amit Gupta, chief AI strategy officer, senior vice president and general manager of Siemens EDA, stated that the Fuse EDA AI Agent evolves the Fuse EDA AI system from in-tool capabilities to autonomous end-to-end workflow orchestration, enabling reduced design cycles while maintaining quality standards. The open architecture supports customer integration of custom workflows and models for enterprise-scale deployment.
Jung Yun Choi, executive vice president of Memory Design Technology at Samsung Electronics, noted that Samsung is introducing Siemens' Fuse as a key enabler for agentic semiconductor workflows, with its architecture expected to accelerate progress beyond traditional automation and enhance engineering productivity and design outcomes.
