Electronicseng Semiconductors: Agentic EDA for 3D IC Electrothermal Co-Design

For stacked-die semiconductor programs, the practical constraint is moving electrical, power, thermal and package data through a single design space before late signoff. A recent update connects the agentic AI capabilities introduced at DAC 2026 with NVIDIA accelerated computing, so teams can evaluate 3D IC electrothermal behavior across seven EDA tools and test hundreds of design points per cycle.

Semiconductor framing: why electrothermal co-design matters

AI accelerators demand higher compute capacity, wider memory bandwidth and tighter thermal control inside compact packages. The NVIDIA Rubin GPU, for example, supports up to 50 petaflops of NVFP4 inference and up to 288 GB of HBM4 per GPU; those GPUs can be combined into systems such as NVIDIA Vera Rubin NVL72. As compute and memory move closer together, power density and heat concentration increase. The implication for 3D IC development is that electrothermal co-design belongs in design loops, not only in final verification.

Coordination layer and benchmark claims

At the center of the release is a coordination layer that aligns Siemens electrothermal analysis tools through one workflow. The Siemens Design Space Exploration and Optimization (DSE/DSO) technology lets engineers test and validate alternatives with physics-based signoff solvers. The benchmark data cited here comes from Siemens reported studies: up to a 95% reduction in engineering effort, up to a 99% reduction in workflow setup time, and up to 50 times greater design exploration coverage compared with manual workflows.

Comparison: manual coordination vs agentic workflow

Workflow aspect	Manual coordination pattern	Agentic coordination pattern
Tool coordination	Specialists coordinate multiple tools and transfer data across electrical, power, thermal and extraction analyses.	A coordination layer aligns Siemens electrothermal analysis tools through one workflow; the Fuse EDA AI Agent, integrated with Intelligence Center X, coordinates multiple Siemens tools.
Evaluation pattern	Teams may reduce the number of scenarios evaluated under time and computing constraints.	The agent runs coupled electrothermal analyses, guides design-space exploration, assesses trade-offs and reruns selected configurations against objectives.
Issue discovery	Power and thermal issues can be identified later.	It helps engineers evaluate additional alternatives and identify power and thermal issues earlier in development.
Scale	Manual effort is the comparison baseline.	Using seven EDA tools, the workflow coordinates electrothermal analysis and evaluates hundreds of design points per cycle.

Agentic workflow anatomy

The engineer defines design objectives. The Fuse EDA AI Agent, integrated with Intelligence Center X, coordinates multiple Siemens tools. It runs coupled electrothermal analyses, finds unmet targets, and evaluates whether causes involve power delivery, placement, materials or a combination of factors. Based on the results, it guides design-space exploration, assesses trade-offs and reruns selected configurations against the objectives.

The workflow coordinates nine steps across seven Siemens EDA tools. It performs 50 to 100 reasoning steps during an eight- to 10-hour runtime and can evaluate heat transfer coefficient (HTC) boundaries, power budgets and material stack-ups across 200 to more than 500 design points per cycle.

Checklist: Siemens EDA tool chain

1. Innovator3D IC Integrator software: supports floor planning and 3D IC assembly and package design.
2. Solido Characterizer software: characterizes libraries to generate Liberty files.
3. mPower Digital software: performs power analysis and calculates temperature-dependent instance power.
4. Calibre 3DThermal software: supports 3D IC thermal signoff and power mapping.
5. Simcenter Flotherm software: provides system-level thermal simulation and visualization.
6. Calibre xACT software: performs temperature-dependent parasitic extraction.
7. Solido SPICE software: supports power- and temperature-aware electrical SPICE simulation.

NVIDIA technologies used in the workflow

Siemens combines the NVIDIA Agent Toolkit with NVIDIA accelerated computing and CUDA-X libraries to support the agent-driven EDA workflow. Accelerated simulation tools supply analysis results that agents evaluate during design exploration.

• Nemotron open models: support the orchestration and execution of agentic EDA subflows.
• NeMo Gym: tracks agent trajectories and identifies areas for refining agent skills and Model Context Protocol (MCP) tool descriptions.
• NeMo RL: supports post-training fine-tuning of language models with EDA knowledge and tool-calling capabilities.
• NeMo Relay: provides observability and audit records across agent harnesses.
• NeMo Switchyard: routes tasks based on compute requirements, accuracy and token costs.
• NVIDIA OpenShell: sandboxes agent execution to help protect intellectual property and foundry process design kits (PDKs).
• NVIDIA CUDA-X libraries: accelerate supported engineering solvers used in the workflow to reduce simulation time during iterative design.

Common misconceptions

• The workflow is not a substitute for physics-based signoff solvers; DSE/DSO is described as enabling engineers to test and validate alternatives with them.
• It is not limited to thermal checking alone; the coordinated analyses cover electrical, power, thermal and package-related exploration.
• Earlier issue detection does not mean removing engineer judgment; the engineer defines the design objectives, and the agent evaluates causes such as power delivery, placement, materials or combinations.

Action suggestions for semiconductor teams

1. When planning a 3D IC program, place electrothermal co-design inside the design loop rather than treating it as final verification only.
2. Review the tool chain inventory across seven EDA tools, especially Innovator3D IC Integrator software, Calibre 3DThermal software and Simcenter Flotherm software, to identify hand-off points.
3. Use the reported benchmark scope as a comparison baseline: up to a 95% reduction in engineering effort, up to a 99% reduction in workflow setup time, and up to 50 times greater design exploration coverage compared with manual workflows.
4. Protect process and IP data when evaluating agent execution; NVIDIA OpenShell is described as sandboxing agent execution to help protect intellectual property and foundry process design kits (PDKs).

FAQ

What does the update combine?

It combines agentic AI capabilities introduced at DAC 2026 with NVIDIA accelerated computing to support electrical, power, thermal and co-design analysis for 3D integrated circuits (ICs).

How many tools and steps are coordinated?

The workflow coordinates nine steps across seven Siemens EDA tools and uses seven EDA tools for electrothermal analysis.

What scale of exploration is described?

It can evaluate heat transfer coefficient (HTC) boundaries, power budgets and material stack-ups across 200 to more than 500 design points per cycle, and it evaluates hundreds of design points per cycle in the broader workflow description.

Which AI accelerator capability is cited?

The cited example is the NVIDIA Rubin GPU, which supports up to 50 petaflops of NVFP4 inference and up to 288 GB of HBM4 per GPU.