Synopsys and Ansys power the future of innovation—connecting silicon to systems.
EDA
System
IP
SolvNetPlus
Training & Education
A new wave of intelligent machines will soon be unleashed: robots and other physical AI systems that can perceive their environment, make decisions on the fly, and act autonomously.
“AI is already expanding beyond the data center and increasingly finding its way into edge form factors,” says Olivier Blanchard, research director and AI devices lead at The Futurum Group.
That expansion is expected to accelerate in the years ahead. According to The Futurum Group, the total addressable market for robotics will grow from $83 billion in 2025 to $391 billion by 2035.
“The acceleration in R&D funding that lays the groundwork for the next wave of AI-driven disruption — one that will usher the deployment, at scale, of complex mobile and humanoid robots — is already well underway,” Blanchard says.
Turning that investment into deployable physical AI systems requires a new form of engineering: holistic silicon-to-system design, validation, and physics-aware training. Because system behavior depends on the interaction of silicon, sensors, software, mechanics, and power, those elements must be co-designed and validated together to help ensure safe, secure, and reliable operation. Just as important, the AI systems controlling those machines need training data grounded in physical reality.
Applying automotive discipline to physical AI
The automotive industry offers insight into the engineering rigor required to develop physical AI systems. Many modern vehicles already sense their surroundings, process data locally, make real-time decisions, defend against cyber threats, and meet stringent safety requirements.
As robots, drones, industrial machines, and other forms of physical AI take on more autonomous work, they will need to deliver similar capabilities.
“We don’t really separate a robot from a car,” says Anis Jarrar, senior fellow at NXP Semiconductors, a global semiconductor company developing chips for physical AI and automotive applications. “Basically, a car is a robot on wheels.”
Safety and predictability are paramount for any autonomous system, he says, and they must be established long before anything is physically built. That puts pressure on engineering teams to validate chips and software earlier in the design process.
“Part of it is leveraging emulation and verification — pre-silicon — so that everything is predictable,” Jarrar says. “Proving the functionality of the chip and validating the software before the silicon ships out has been a tremendous step in accelerating our time to market.”
Modeling, training, and transferring physical AI to the real world
Chips are only one part of the equation. The machines, their functions, the environments in which they will operate, and the conditions they will encounter must also be modeled, simulated, and validated.
That becomes especially important for tasks that require dexterity, object handling, and fine motor control. Connecting or disconnecting a cable, for example, may be routine for a human, but it’s a tall task for a robot.
Cables bend, connectors resist, and objects behave differently depending on force, angle, material, and contact. A robot may need to interpret sensor data, adjust its grip, and respond to small physical changes as a task unfolds.
Preparing robots for real-world deployment requires training and validation across an enormous range of physical interactions and operating conditions. Physics-aware synthetic data generated from high-fidelity simulations helps expose AI models to those scenarios before deployment.
Analog Devices, a global semiconductor company with deep expertise in sensing and edge intelligence, is working with Synopsys and NVIDIA to build the simulation foundation for that training. The collaboration combines system-level modeling, virtual prototyping, multiphysics simulation, synthetic data generation, and training environments for industrial robotics.
“Synopsys’ multiphysics simulation is a critical enabler of realistic robotic test benches,” says Paul Golding, vice president of Edge AI at Analog Devices. “Together with NVIDIA, we’re using that fidelity to create benchmarks and digital twins that make sim-to-real transfer practical for real industrial dexterity.”
Schaeffler, a global motion technology company serving automotive and industrial markets, is also using Synopsys simulation capabilities, including Ansys solutions, to evaluate humanoid robots and their components before industrial deployment. This includes modeling motion, energy use, mechanical stress, sensor feedback, safety features, and the physical conditions that influence system behavior.
“The future of humanoids depends on production readiness, uptime, and serviceability,” says David Kehr, president of Schaeffler’s Humanoid Robotics division.
Digital twins and simulation can support the entire product lifecycle, he adds, from early validation to component wear, diagnostics, software updates, and performance improvements after deployment.
Preparing physical AI for deployment
The opportunity for physical AI is significant, but real-world deployment requires hard engineering work. Simulation and digital twins are pulling that work into earlier stages of the engineering process, and they are stretching co-design from silicon to system.
Teams can test chip behavior before silicon is available, generate physics-aware synthetic data to train robotics systems in realistic virtual environments, evaluate physical interactions before hardware testing, and use digital models to support post-deployment diagnostics, maintenance, and software updates.
Together, these capabilities are helping turn physical AI from a promising research field into deployable machines that can operate in complex real-world environments.
Continue Reading
Building India's Next Wave of Semiconductor Innovation: Synopsys at SEMICON India 2026
Architecting AI Factories: Rack-Scale Design Series
