Chipdesign og verifikation med AI-agent
Siemens accelererer design og verifikation af integrerede kredsløb med AI-agent i firmaets Questa One softwareportefølje (in english).
Siemens has announced the Questa One Agentic Toolkit, which brings domain-scoped agentic AI workflows to its Questa One smart verification software portfolio to accelerate creation, verification planning, execution, debugging and closure to achieve trusted RTL sign-off faster while transforming how engineers approach integrated circuit (IC) design and verification tasks.
The verification productivity gap continues to widen as design complexity explodes with 3D ICs, chiplet-based architectures and software-defined systems.
The Questa One Agentic Toolkit transforms verification and design from isolated tool interactions into intelligent, domain-scoped multi-step, framework-agnostic workflows powered by agentic AI - autonomous systems operating within the verification domain under customer-defined governance boundaries that can reason, plan and execute complex tasks while maintaining configurable human oversight at critical decision points directly within engineers' existing environments.
- Questa One sets a new standard for smart verification and the Questa One Agentic Toolkit builds on those connected and data-driven principles with agentic AI workflows that empower our customers to achieve trusted design and verification closure with AI acceleration – while maintaining the human expertise and judgment that builds quality and trust, says Abhi Kolpekwar, senior vice president and general manager, Digital Verification Technologies, Siemens Digital Industries Software.
- We own the engines and we understand verification at the deepest level. It isn’t an afterthought or a wrapper; it’s the EDA industry’s most comprehensive verification solution enhanced with intelligent agentic AI workflows that work the way our customers expect and delivers the results they are demanding.
Working with Fuse EDA AI system and beyond
The Questa One Agentic Toolkit works seamlessly with the Fuse EDA AI system, Siemens’ agentic and generative framework for electronic design automation, providing customers who want a fully integrated Siemens experience with optimized performance and deep integration.
It is Fuse-preferred, with enhanced capabilities when used within the Fuse environment. Recognizing that customer choice is paramount, the framework-agnostic architecture protects existing investments and integrates with other agentic platforms without compromise. Whether teams use existing or emerging frameworks, Questa One agentic workflows adapt to their environment with standardized interfaces that work consistently across platforms.
The Siemens advantage: Engines, integration and openness
These intelligent agentic workflows represent a fundamentally different approach than what's emerging from startups and point-solution providers. Siemens uniquely combines verification engine expertise with deep AI integration and customer choice through three differentiating pillars:
Engine-native intelligence: Siemens creates both the industry-leading Questa One tools and the model context protocols (MCPs) that expose them to agentic frameworks. These workflows, built leveraging NVIDIA Llama Nemotron and NVIDIA NIM, understand verification state in real time and maintain comprehensive awareness and contextual intelligence relationships between designs, testbenches, test plans and specifications. This provides customers with autonomous goal decomposition, adaptive cross-run strategies and persistent expertise building.
Coding application and platform agnostic: The solution works with main-stream AI coding applications - including GitHub Copilot, Claude Code, Cursor, Cline and Siemens' own Fuse - and can be used in CLI mode or IDEs (e.g. VS-Code). While optimized as Fuse-preferred for customers who want a fully integrated experience based on Siemens’ toolset, these workflows remain completely agnostic, adapting to customer workflows rather than forcing adaptation.
Scalable, connected, data-driven foundation: By leveraging the connected verification ecosystem that dynamically orchestrates between tools, including Questa One, Tessent software for DFT and the Veloce CS hardware-assisted verification and validation system, these agentic workflows bring AI-powered capabilities across the entire design and verification landscape.
Intelligent workflows across design and verification
The Questa One Agentic Toolkit launches with the following intelligent agents that demonstrate the potential of agentic AI:
The RTL Code Agent generates synthesizable RTL code from natural language descriptions while simultaneously checking for coding violations and suggesting fixes aligned with industry standards, presenting engineers with clean, high-quality RTL for review.
The Lint Agent optimally configures lint analysis, reading existing RTL code to check for design errors and coding style violations. Designers then review results and are offered automated AI-powered design fixes or waivers to help to ensure the highest-quality RTL.
The CDC Agent optimally configures, then runs clock domain crossing verification on a design. Configuration fine-tuning suggestions are made based on results. This empowers designers to achieve the cleanest asynchronous designs after reviewing the results and enabling automated AI-powered design fixes or waivers.
The Verification Planning Agent analyzes design specifications and automatically generates comprehensive verification plans. Engineers review and approve each step while the AI handles structuring sections, creating detailed feature descriptions and defining scenarios and checking strategies.
The Debug Agent accelerates root cause analysis by intelligently correlating waveforms, assertions, coverage data and log files. It identifies suspicious signal transitions, suggests potential failure mechanisms and generates targeted debug scenarios for engineer reviews.
These agents leverage the toolkit's MCPs to work directly with Questa One Verification IQ, Questa One SFV, Questa One Sim and other tools, backed by curated prompt libraries developed by domain experts.
