Cadence pairs AI with Nvidia, debuts layout agent on Google Cloud

At CadenceLIVE this week, Cadence Design Systems announced it will link its multiphysics simulation and system-design tools with Nvidia’s AI stack and Omniverse, and will roll out a chip-layout AI agent on Google Cloud that uses Google’s Gemini models.

The Nvidia integration combines Cadence thermal, electrical and mechanical simulation engines with Nvidia CUDA-X libraries, AI models and the Omniverse simulation environment. Engineers can model how chips, cooling, power and networking interact and run virtual tests before hardware is built or deployed.

The effort targets robotics and system-level design by connecting Cadence physics-based models to Nvidia models used to train robots in simulated environments. Training in simulation reduces the need to collect large datasets from physical robots, and the companies noted that those synthetic datasets must be generated with accurate physics models.

Cadence CEO Anirudh Devgan noted, “The more accurate generated training data is, the better the model will be.” Nvidia CEO Jensen Huang told the audience the firms are “working with you across the board on robotic systems.”

Nvidia highlighted industrial users of its Isaac framework and Omniverse tools, including ABB Robotics, FANUC, YASKAWA and KUKA, which use virtual commissioning to test production systems in software before physical rollout. The partners noted digital twin testing can model entire production lines and complex robot operations in physically accurate virtual settings.

Separately, Cadence introduced an AI agent focused on physical layout that translates circuit designs into silicon implementations. The agent will be offered through Google Cloud and pairs Cadence electronic design automation tools with Google’s Gemini models.

The layout agent is part of Cadence’s ChipStack AI Super Agent platform, which uses model-based reasoning with native design tools to coordinate tasks across design stages. Cadence reported up to 10 times productivity gains in early deployments across design and verification tasks and did not disclose customers.

Cloud delivery lets design teams run compute-heavy layout and verification workloads without on-premise infrastructure. Cadence stated the ChipStack agent can interpret design requirements and perform tasks across stages to speed handoffs between front-end design and back-end implementation.

At the event, Nvidia introduced NVIDIA Ising, a set of open-source quantum AI models intended for quantum processor calibration and quantum error correction. Nvidia reported up to 2.5 times faster performance and three times higher accuracy in decoding processes used for error correction.

The companies emphasized that system performance depends on interactions among compute, networking, cooling and power subsystems and noted digital twin testing allows engineers to evaluate tradeoffs and optimize configurations before building hardware.