Synopsys and Ansys power the future of innovation—connecting silicon to systems.
EDA
System
IP
SolvNetPlus
Training & Education
When asked what concerns him most about the state of the semiconductor industry, Synopsys President and CEO Sassine Ghazi didn’t hesitate.
“The pace, the pace, the pace,” he said.
The answer came during a fireside chat at SEMICON West with Ed Sperling, editor-in-chief of Semiconductor Engineering. The two covered autonomous agents, 3DICs, digital twins, and physical AI, and each topic came back to the same pressure point: AI is changing both the chips the industry builds and the way engineers build them, faster than traditional development cycles can handle.
“I’ve never seen a pace and system-level co-design and complexity like what we’re dealing with right now,” Ghazi said.
Enhance Your Chip Design with AI
Explore the enhanced Synopsys.ai brochure, featuring cutting-edge advancements in Advanced Optimization, Generative AI, and Agentic AI to transform your chip design process.
From prompts to outcomes
To address these challenges, Ghazi said Synopsys has been working to “re-engineer engineering.” Individual components of a product can no longer be designed and optimized in isolation, he explained. Instead, products must be engineered from the very beginning as a unified whole.
That requires co-design across software, chiplets, the package, and underlying process technology. And thermal, mechanical, electromagnetic, and power effects must be evaluated when architectural decisions are being made, not discovered after implementation.
AI will be essential for managing this complexity, Ghazi noted, and it is becoming increasingly capable.
“We started from an assistant copilot, to generative capability, to now what we’re describing as long-horizon autonomous agents,” Ghazi said. “Instead of giving a prompt, you give an outcome.”
This shift changes the role of AI from answering questions or optimizing a single task to orchestrating many specialized engineering activities. A long-horizon agent might interpret an objective, call upon multiple task agents, use proven design tools, evaluate results, explore alternatives, and advance the design toward a verified outcome.
That does not remove engineers from the process, Ghazi emphasized. It elevates their role, accelerates their work, and expands their capacity.
“You still need the human judgment to define the architecture, to guide, and to sign off the final check before you go into product manufacturing,” he said.
Bringing physics into system design
In addition to playing a greater role in engineering workflows, AI is moving into vehicles, robots, drones, and other physical systems. Designing these products requires engineers to account for how electronics, software, and mechanical components interact with one another and with the surrounding environment.
Sperling characterized this as a “system of heterogeneous systems.” Ghazi agreed, noting that even the meaning of “system” shifts with scale. An advanced multi-die package is itself a system, but it becomes a component within a rack, and a rack becomes a subsystem within a data center. At each level, thermal, mechanical, electromagnetic, and power effects can create performance, reliability, and manufacturability challenges.
3DICs offer a clear example. Recalling predictions that 2026 would be the year of the 3DIC, Sperling asked: “Is it real?”
Many 3DICs are already running in data centers, Ghazi said, where demanding workloads generate heat that can warp or crack chiplets. Those effects must be understood and accounted for before the architecture is set. The ability to provide multiphysics signoff, he added, was central to Synopsys’ acquisition of Ansys.
Digital twins extend multiphysics simulation and analysis to physical AI systems and their operating environments. Many companies already use digital twins within individual engineering domains, from virtual prototypes of chips to digital crash testing of vehicles. The next step is to connect electronics, software, physics, and the environment within a shared virtual model.
“I’m not sure how you can build a future sophisticated intelligent system without having a digital twin,” Ghazi said. “It will take you forever, and the cost will be astronomical.”
Left to right: Sassine Ghazi (Synopsys) and Ed Sperling (Semiconductor Engineering)
Racing the model
AI models change every three to four months, Ghazi said, while silicon has traditionally been updated every two to three years. System companies now want spec-to-tape-out cycles shorter than nine months. They’re also comfortable with chips that serve for only two or three years.
Keeping up means changing workflows while still meeting current commitments.
“It’s very easy to put it on a PowerPoint slide,” Ghazi said. “It’s very hard for a company to deliver the current commitment to their customer and transform that same engineering workforce.”
Managing the pace Ghazi described will require engineers to move faster without lowering the standard for what can be trusted. AI can explore more options and compress development cycles, but its results will only be as reliable as the models, engineering tools, and physics-based solvers behind them.
Read Ed Sperling’s Q&A with Sassine Ghazi in Semiconductor Engineering
Continue Reading
When Everything Interacts: Engineering Intelligent Systems in a Multiphysics World
