Robots trained inside computer simulations have a persistent habit of stumbling when they meet the real world. The physics never quite match — surfaces behave differently, objects flex in unexpected ways, lighting throws off sensors. That gap between virtual performance and physical competence has been one of industrial robotics' most stubborn bottlenecks, and two of Silicon Valley's biggest names are now throwing their combined engineering weight at eliminating it.
At CadenceLIVE Silicon Valley 2026 on April 16, Cadence Design Systems and NVIDIA announced an expanded partnership that integrates Cadence's high-fidelity multiphysics simulation engines with NVIDIA's Isaac open-source simulation libraries, Cosmos open-world models, and Jetson edge AI hardware. The result is an end-to-end pipeline that spans from synthetic world-model training through physics validation to real-world deployment — all coordinated by AI agents across the full lifecycle.
Bridging the Chasm
The sim-to-real gap is not a new problem, but it has taken on new urgency. As companies race to deploy physical AI systems in manufacturing, logistics, and energy, the quality of simulation training data has become a critical constraint. Robots trained on approximate physics learn approximate behaviors, which often means they fail on the factory floor — sometimes expensively.
Cadence, best known as a dominant supplier of electronic design automation software, also makes physics engines that model real-world material interactions with exacting precision: how metals deform, how fluids flow, how surfaces make contact. These capabilities, long used in aerospace and semiconductor design, are now being pointed at a different target — generating the training data that robot AI systems need.
"The more accurate the generated training data is, the better the model will be," Cadence CEO Anirudh Devgan said at the Santa Clara conference. "Our expanded collaboration with NVIDIA accelerates the convergence of design and physical realization, connecting the Cadence AgentStack, Physical AI Stack, and AI factory digital twins with NVIDIA's breakthroughs in accelerated computing to deliver unprecedented speed, accuracy and trust in simulation and system development."
NVIDIA CEO Jensen Huang framed the ambition in broader terms. "We are at an inflection point in computing — CUDA-accelerated computing and AI are reinventing the engineering process," Huang said. "For the first time, we can innovate in the digital world — exploring, testing, and optimizing ideas at unprecedented speed and scale — by building everything as full-fidelity digital twins first."
How the Stack Works
The joint platform coordinates AI agents across the full robotics development lifecycle. Virtual training runs inside NVIDIA Isaac Sim and Isaac Lab, where robots learn manipulation and navigation tasks using synthetic data generated by Cosmos world models. Those policies are then evaluated against Cadence's detailed multiphysics solvers and tested at mission scale in VTD and VTDx, Cadence's extended simulation environments for complex real-world scenarios.
Once validated, models deploy to NVIDIA Jetson robotics hardware at the edge. A live virtual twin continues to run alongside the physical system, enabling continuous monitoring and refinement. The key innovation is that accurate physics is embedded at every stage — training, validation, and inference — rather than bolted on as an afterthought.
Cadence says the integration will deliver engineering workflows up to 100 times faster than current approaches, with AI physics models accelerating its principled solvers across both electronic design and physical AI applications.
A Crowded Race
The partnership does not exist in isolation. NVIDIA has been systematically building simulation alliances across the industrial engineering landscape. In March 2026, ABB Robotics announced that its RobotStudio HyperReality platform, powered by NVIDIA Omniverse, had achieved 99 percent sim-to-real accuracy in early pilots with Foxconn and other manufacturers. ABB reported that the technology could cut robot setup and commissioning times by up to 80 percent and reduce deployment costs by 40 percent. NVIDIA has also forged similar ties with Siemens and Dassault Systemes for industrial digital twins.
What Cadence brings is depth in multiphysics fidelity. While Omniverse provides the rendering and scene-generation backbone, Cadence's solvers model the fine-grained material interactions — thermal, electromagnetic, structural, fluid — that determine whether a robotic gripper can reliably handle a deformable object or whether a welding arm applies consistent force. In robotics applications where millimeter-level precision matters, that physics resolution can be the difference between a simulation that transfers cleanly and one that does not.
What It Means
For Cadence, the robotics angle represents a significant strategic expansion. The company's 2024 revenue hit $4.6 billion, almost entirely from chip and system design tools. Moving into the AI infrastructure layer — providing the physics backbone for robot training pipelines — opens a large and growing addressable market as physical AI moves from research labs to production lines.
For NVIDIA, Cadence joins a roster of simulation partners that collectively aim to make the company's hardware the default deployment target for industrial robotics. Every robot trained inside the NVIDIA simulation stack is a robot that ships with a Jetson module inside it.
The broader signal is that the robotics industry is converging on a shared architectural pattern: train in simulation, validate with high-fidelity physics, deploy on accelerated edge hardware, and keep a digital twin running for continuous improvement. The companies that can deliver each link in that chain with production-grade reliability stand to define how the next generation of industrial robots reaches the factory floor.
Early customers and partners — including Ascendence, Argonne National Laboratory, Honda R&D, Samsung, and SK Hynix — are already testing Cadence solutions accelerated by NVIDIA hardware. NVIDIA itself is adopting Cadence's AgentStack in its own semiconductor and system design flows, providing real-world feedback to harden the platform for broader deployment.
The sim-to-real gap has stalled industrial robotics for years. With physics simulation, AI training, and edge deployment now stitched into a single agent-orchestrated workflow, Cadence and NVIDIA are betting that gap is about to close for good.
“We are at an inflection point in computing.”— Jensen Huang, Founder and CEO, NVIDIA