EDA
System
IP
SolvNetPlus
Training & Education
As intelligent systems proliferate, the nature of engineering itself is changing.
What once could be designed, validated, and optimized within largely separate domains now spans extraordinary ranges of scale, physics, and time — along with the boundary conditions that emerge at their intersections.
With that shift comes a clear mandate: engineering organizations must rethink how they design, simulate, and deliver complex systems.
That message came through clearly during a panel discussion at Synopsys Converge 2026, where leaders from Boeing, Mercedes‑Benz, Meta, and Synopsys compared notes on the realities of building and operating next-generation products.
Despite working in vastly different markets, the panelists described a shared challenge: managing complexity that no single discipline, tool, or timeframe can fully capture on its own.
Multiphysics Fusion Technology for Multi-Die Designs Explained
Learn why multiphysics analysis must move earlier in the design flow.
Confronting complexity and scale
AI-powered glasses were a telling example. At Meta, engineers are creating systems in which form factor, sensing, and compute all collide.
“We’re creating a wearable compute platform in a very small form factor,” said Jean Boufarhat, vice president and head of silicon for Meta’s Reality Labs. “Our glasses must integrate optics, displays, chips, materials, and style.”
What makes that integration especially difficult is the sheer range of scales involved.
“We operate across angstroms in chip design, millimeters in devices, meters in manufacturing equipment, and kilometers for data centers,” Boufarhat said. “And we also operate across time scales — from nanoseconds for chip behavior to years for durability.”
Engineering teams are also running up against the laws of physics.
“In autonomous driving, the complexity is truly multiphysics — optics, thermals, electromagnetism, voltage fluctuations, sonar,” said Sundar Ramalingam, vice president of research and development at Mercedes-Benz North America. “Everything interacts.”
These interactions make it increasingly difficult to isolate decisions or defer validation until late in the development process. System-level decisions must now be modeled holistically and earlier than ever.
Defining silicon footprints
Ravi Subramanian, chief product management officer at Synopsys, framed the challenge more broadly. Across industries, he sees three forces converging.
“Products are becoming software-defined and therefore silicon-powered,” he said. “Innovation now happens continuously, even post-sale. And product development costs need to come down.”
That combination fundamentally changes how systems must be designed.
Ramalingam said Mercedes-Benz has historically made silicon decisions four years ahead of production, but rapid changes in software — and especially AI — has introduced new levels of uncertainty and risk.
“Traditionally, software followed silicon,” he noted. “With the assimilation of LLMs, we are finding it extremely hard to define what kind of silicon footprint we need.”
Instead of locking onto a single, fixed footprint, Mercedes-Benz is now considering architectural approaches that provide more flexibility.
“Do we choose a different kind of design architecture that allows us to replace chips or chiplets in the progression of the lifetime of the vehicle?” Ramalingam asked rhetorically.
Regardless of the answer, he said one thing has become evident: “Compute should follow software.”
Embracing digital twins and AI
Throughout the panel, leaders emphasized that traditional build-and-test approaches are no longer viable — economically or technically.
At Boeing, engineering teams are embracing digital tools to address complexity earlier in the product development process.
“Advances in simulation and digital twins let us move more of the design into the virtual world,” said Todd Citron, the company’s CTO. “A lot of this is about ‘shifting left,’ bringing in considerations that historically would be well downstream.”
AI is deeply embedded in the workflows, he added, with LLM copilots trained on Boeing data.
“We have 160,000 users and 38,000 active users per month,” said Citron. “The next step is agentic AI — automating entire workflows, not just assisting tasks.”
At Meta, AI has become both a productivity tool and a democratizing force.
“AI makes previously expert-only tools accessible,” Boufarhat said. “It turns managers or non-experts into builders.”
Turning human experience into code
Asked how their organizations might look three years from now, the panelists were cautious but clear about the path ahead.
“Small teams with agentic AI will build what once required large teams,” Boufarhat predicted. “Physics will remain the ultimate bottleneck. But we must accelerate everything up to that point.”
Citron agreed that workflow automation would be transformational but emphasized the human side of the transition.
“The challenge is training teams to use these tools and turning human expertise into code,” he said.
Subramanian closed with a note of urgency — and optimism.
“Everything will happen faster and be more disruptive,” he said. “Courage will be required to embrace the change.”
For organizations willing to rethink workflows, invest in holistic modeling, and empower engineers to work alongside intelligent agents, the opportunity is enormous. For those that don’t, complexity will continue to surface where it’s hardest and most expensive to fix: in the real world.
Continue Reading
From Overdesign to Co-Design: Confronting Multiphysics Challenges in Chips
Taming Advanced Node Physical Verification – NVIDIA’s Perspective
