How Simulation Is Driving Physical AI for Data Centers

ADI, Synopsys, and NVIDIA demonstrate robotics cable management benchmarks for data center applications at NVIDIA GTC.

If data is the new oil, then cables are the pipelines. The demand for AI-based services has led to an explosion of hyperscale data centers — facilities designed support thousands of servers that can handle enormous computing, storage, and networking demands. These facilities can encompass tens of thousands of square feet and miles of connecting cables.

At that scale, small inefficiencies quickly add up to big losses. That’s where industrial robotics come into play. However, cables are notoriously challenging to handle — just ask anyone trying to untangle their wired headphone cord before an important call. Even tasks that are second nature to humans, such as plugging in an RJ45 Ethernet cable, are incredibly nuanced when training a robot to do so.

“Nothing is second nature to a robot,” says Kishor Ramaswamy, director of application engineering for Ansys, part of Synopsys. “Which jack should it be plugged into? How should the plug be lined up to the interface? How hard should the cable be held or the connector be pressed to ensure it’s plugged in but not so hard it breaks? What if the cable is caught on something?”

Now scale those connections to thousands of servers in environments that are affected by heat and high-voltage areas, and you’ll get an idea of the challenges faced by robotics companies tackling the cabling conundrum. The explosion of data centers requires quick solutions that only simulation can provide. To meet demand, there isn’t time to train robots in the real world, or for multiple rounds of physical prototypes. Robots need to be trained on synthetic data against standard benchmarks that create a level playing field for evaluating robotic performance. Such physical and virtual benchmarks will help get high-quality products to market in time to capitalize on the groundswell of data center construction.

Analog Devices Inc. (ADI) is up to the challenge. The company is a global semiconductor leader that bridges the physical and digital worlds. Its industrial robotics innovations span motion control, functional safety, advanced sensing, time of flight (ToF), machine learning, high-speed Ethernet communications, and system-level design.

Simulation Provides Speed and Accuracy

ADI is using Ansys Mechanical finite element analysis software to augment the fidelity of cables and connector models in the NVIDIA Isaac Sim environment for training and validating various robotic policies for cable management. Isaac Sim is an open-source robotics simulation framework that enables developers to simulate and test AI-driven robotics solutions in physically based virtual environments.

Benchmarking Robotic Dexterity

ADI plans to use the simulation results to advance its robotics solutions, including ToF sensors and tactile sensors, which provide precise force and torque sensing and measurement. However, the company is also sharing the knowledge it gains as a series of benchmarks to help advance the robotics industry as a whole.

“Synopsys’ multiphysics simulation is a critical enabler of realistic robotic test benches,” said Paul Golding, Vice President of Edge AI at ADI in a press release. “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.”

ADI will provide high-fidelity digital twins of these benchmarks, enabling customers to begin developing AI solutions immediately in NVIDIA Isaac Lab using their own robotic platforms — or by leveraging ADI’s reference design and its high-fidelity digital twin with ADI technology. NVIDIA Isaac Lab is an open-source, unified framework for robot learning. The framework covers everything from environment setup to policy training and supports both imitation and reinforcement learning methods.

ADI’s platform will enable early adopters, including Kawasaki Heavy Industries, to simulate robotic performance and generate synthetic data with greater predictive accuracy, reducing iterative physical testing cycles and accelerating innovation.

Synopsys’ booth at the NVIDIA GTC 2026 March 16-19 in San Jose, California will feature a demonstration of the ADI collaboration, including a bi-manual robotic arm setup with force, vision, and contact sensing, and a corresponding digital twin.

Learn more about Ansys Mechanical software here.