New-Tech Europe | Q3 2026 | Digital Edition

"

frameworks. Agents must manage long-running verification jobs, integrate with legacy schedulers (such as LSF), and handle terabyte-scale datasets in-place. 3. Scalability Across Fragmented Workflows: The vast EDA tool ecosystem can easily overwhelm standard AI, leading to “context saturation” and hallucinations. A unified orchestration layer is essential for deterministic execution across expanding tool chains. Critically, because EDA workflows are inherently multi-vendor and cannot be confined to a specific ecosystem, AI agents require high flexibility to operate seamlessly across diverse toolsets. 4. Opaque EDA Modalities: EDA data involves dense binary formats and opaque databases. Agents require domain-specific parsers to extract actionable intelligence from complex artifacts like waveforms and netlists. 5. Embedded Enterprise Security: To safeguard sensitive IP, agents must operate in highly secure environments. This requires robust Role-Based Access Controls (RBAC), strict sandboxing, comprehensive audit trails, and human-in-the loop checkpoints. The Solution: Siemens’ Fuse EDA AI System and Fuse EDA AI Agent To address these highly specific industry bottlenecks, Siemens has introduced the Fuse™ EDA AI System and Fuse™ EDA AI Agent. This system fundamentally transforms chip and PCB design by integrating generative and agentic AI capabilities across the complete Siemens EDA portfolio. Through a context-aware natural language interface, it intelligently orchestrates complex, multi-tool workflows from initial concept through manufacturing sign-off, increasing engineering productivity and design quality. The architecture is purpose-built for semiconductor engineering. At its core is a centralized, multimodal EDA data lake that uses specialized parsers to break down team and tool silos, ensuring every workflow operates from a single source of truth. Layered on top is an advanced RAG

Once hundreds of these automated sub-flows are developed, they can be strung together to construct comprehensive, multi-tool workflows that span the entire EDA lifecycle. It is much like taking individual Lego pieces to build increasingly larger, modular structures.

"

framework trained on Siemens EDA tools and methodologies, enabling the system to answer complex domain-specific queries with precision rather than approximation. The system is also model-agnostic and open by design - supporting multiple LLMs and integrating seamlessly with third-party tools, reflecting the reality that production EDA environments are inherently multi-vendor. Security and deployment flexibility are built in from the ground up, with support for both on-premises and cloud environments, native RBAC, and comprehensive audit trails that safeguard IP at every layer. Bringing these capabilities together is the Fuse EDA AI Agent, delivering end-to-end automation by planning, orchestrating, and executing workflows across the full semiconductor and PCB system design cycle. This is complemented by AI-driven tool automation and natural-language debugging, allowing engineers to express intent in plain language and have the system translate it into precise, executable actions. Fuse EDA AI Agent also solves scalability issues by centralizing tool discovery within a unified operational layer, effectively preventing context saturation as workflows expand. Ultimately, this creates a system that goes beyond basic assistance, operating autonomously and reliably on behalf of engineers at an enterprise scale. The Fuse EDA AI Agent is built on a highly modular philosophy: each sub-flow from the broader EDA workflow is automated in detail. This is accomplished using a combination of the Model Context Protocol (MCP) for executing EDA tools, “Agent Skills” - executable playbooks that encode domain expertise to properly sequence and set up tools with built in validation and guardrails - and specialized EDA parsers within Fuse EDA AI System to extract precise context from complex EDA data formats (e.g., LEF/DEF, GDSII). Once hundreds of these automated sub-flows are developed, they can be strung together to construct comprehensive, multi tool workflows that span the entire EDA lifecycle. It is much like taking individual Lego pieces to build increasingly larger, modular structures.

Source: Siemens Digital Industries Software

New-Tech Magazine Europe l 29