Agentic AI accelerates 3D IC engineering for the AI factory era

By Sudarshan Deo, software R&D Manager for 3D IC, Siemens

What if you could completely rethink what an EDA workflow looks like for the increasingly complex 3D ICs powering the next generation of AI factories? Siemens’ expanded agentic AI portfolio is automating the hardest parts of 3D IC engineering.

At NVIDIA GTC Berlin 2026, Siemens announced an expansion of its EDA agentic AI portfolio, developed in collaboration with NVIDIA. The announcement represents a significant step forward in applying AI-driven automation to one of the semiconductor industry’s most complex challenges: designing advanced 3D IC systems for the next generation of AI computing.

At the center of this innovation is the Fuse EDA AI Agent, integrated with Intelligence Center X. Together, they orchestrate nine automated workflow stages across seven Siemens electronic design automation (EDA) tools, reducing setup effort by up to 95 percent. This automation enables engineering teams to evaluate up to 50 times more design alternatives, helping them address increasingly difficult power and thermal constraints in advanced 3D IC packages.

AI Factories Are Redefining Chip Design Requirements

AI factories are reshaping semiconductor requirements. As AI workloads grow in scale and complexity, the silicon powering these systems must deliver greater compute performance and memory bandwidth while improving thermal efficiency—all within increasingly constrained physical footprints.

As compute and memory are brought closer together in tightly integrated 3D packages, power density rises and heat becomes more concentrated. Thermal behavior can no longer be addressed late in the design cycle; electrothermal co-design must be a primary consideration from the earliest stages of development.

The Growing Challenge of Engineering Complexity

Designing these advanced systems presents a major challenge for engineering teams.

Today, addressing power, thermal, and performance requirements often requires specialists to manually coordinate workflows across multiple tools and design domains. Engineers must continuously move data between electrical, power, thermal, and extraction analyses while repeatedly iterating to achieve design goals.

As 3D IC architectures integrate more chiplets, interconnects, power-delivery networks, and thermal interfaces within a tightly constrained package, the number of interacting design variables grows significantly. Time and compute limitations can restrict how many of these coupled scenarios engineering teams evaluate, allowing power, thermal, or performance bottlenecks to remain hidden until late in the design process, when changes are more costly and difficult to implement.

To find the best solution with confidence, engineering teams need to analyze far more design alternatives than current workflows allow.

Introducing an Agentic Approach to 3D IC Engineering

Siemens is introducing a new engineering model designed to address this challenge, using the Fuse EDA AI Agent, integrated with Intelligence Center X.

Engineers begin by defining a design objective: from there, the AI agent coordinates the workflow across multiple Siemens tools, automating the execution of complex multi-domain analyses.

The agent runs coupled electrothermal simulations, evaluates results, identifies missed design targets, and helps diagnose root causes. Rather than requiring engineers to manually trace issues across multiple domains, the agent can determine whether performance limitations originate from power delivery challenges, placement decisions, material selections, or an interaction among these factors.

This agent-driven approach allows engineers to focus more on innovation and decision-making while reducing the manual effort typically required to manage complex workflows.

Accelerating Exploration with NVIDIA AI Technologies

To support these advanced engineering workflows, Siemens leverages NVIDIA Agent Toolkit alongside NVIDIA accelerated computing platforms and CUDA-X libraries.

This combination enables an AI-driven workflow where accelerated simulation tools provide faster results for the agent to evaluate. By shortening simulation turnaround times, engineers can explore substantially more design possibilities within the same project schedule.

The ability to analyze a far greater number of alternatives is particularly important in advanced 3D IC design, where optimal solutions often emerge only after evaluating numerous trade-offs between performance, power efficiency, thermal behavior, package architecture, and material choices.

From Automation to Engineering Intelligence

Traditional workflows that rely on extensive manual coordination cannot keep pace with the increasing complexity of advanced 3D IC architectures.

By combining Siemens’ EDA expertise with NVIDIA’s AI and accelerated computing technologies, the Fuse EDA AI Agent and Intelligence Center X introduce a new approach to engineering. Rather than simply automating individual tasks, they enable an intelligent, agent-driven workflow capable of orchestrating analyses, diagnosing design challenges, and expanding design exploration at greater scale.

The result is a faster path to optimized designs, allowing engineering teams to address the power and thermal challenges of next-generation AI systems while accelerating innovation for the AI factory era.

See how Siemens is advancing agentic AI for 3D IC engineering, learn more here: Siemens expands agentic AI to tackle 3D IC thermal | Siemens

Sudarshan Deo is a Software R&D Manager for 3D IC at Siemens in the Central Engineering Solutions Division. He holds an M.S. in Computer Science from California State University, Sacramento, and has expertise in implementing and integrating complex 3D IC workflows, including physical design, DFT, multi-physics, thermal, stress, SI, PI, and DC IR-drop analysis. He also contributes to initiatives in 3D UI and data management using IP lifecycle management and integrates advanced AI frameworks to optimize and enhance 3D IC workflows. He serves as a technical point of contact for divisions across Siemens Digital Industries Software. 

Also Read:

Siemens and NVIDIA Accelerate Lithography for Next-Generation AI Chips

The silent killer of analog reliability: Why SPICE misses floating net

Wilson Survey Signals Need for New Understanding of Verification Success