Cadence–NVIDIA AI Flows

Cadence and NVIDIA said on April 15 they are widening their partnership to tie chip-design software more tightly to AI agents, physics simulation and digital twins at CadenceLIVE Silicon Valley 2026. (cadence.com) Cadence announced the move at its April 15-16 event in Santa Clara, California, where Chief Executive Anirudh Devgan was scheduled to appear with NVIDIA Chief Executive Jensen Huang in a fireside chat. (cadence.com) Electronic design automation is the software layer chip companies use to draw circuits, verify them and prepare them for manufacturing, and Cadence said the new work links that layer with NVIDIA CUDA-X, AI physics software and Omniverse digital-twin tools. (cadence.com) In plain terms, the companies are trying to let software agents handle longer stretches of chip and system work while simulation models check whether those designs still obey the laws of heat, power and signal behavior. Cadence said those flows span semiconductors, physical AI systems and hyperscale AI factories. (cadence.com) The push started before CadenceLIVE. On March 17, Cadence said it was expanding its collaboration with NVIDIA for “agentic” integrated-circuit design, and described long-running software agents that can generate designs, debug errors and manage end-to-end workflows. (cadence.com) That March announcement also put numbers on the hardware side: Cadence said its expanded portfolio on NVIDIA Grace central processors and Blackwell graphics processors can deliver up to 80 times greater throughput and up to 20 times lower power consumption, depending on the workload. (cadence.com) NVIDIA has been making the same pitch across industrial software. At GTC on March 16, it said Cadence, Siemens, Synopsys, Dassault Systèmes and PTC were building NVIDIA-powered agents to plan, optimize and verify chip and system workflows. (nvidia.com) For Cadence, that means selling more than a design tool license. The company is packaging AI agents, simulation software and computing infrastructure together, including the Millennium M2000 supercomputer line it introduced with NVIDIA support in 2025. (nvidia.com) NVIDIA said in 2026 that Cadence’s ChipStack AI SuperAgent is aimed at semiconductor design and verification tasks including design coding, testbench coding, test-plan creation and debugging. That is the kind of integration work that usually requires services teams to connect software tools, chip intellectual property blocks and customer-specific design flows. (nvidia.com) Cadence said customers and partners including Ascendence, Argonne National Laboratory, Honda R&D, Samsung and SK hynix are already using Cadence software accelerated by NVIDIA. The companies are betting that faster simulation and more automated design loops will pull AI deeper into the day-to-day work of building chips and the data centers that run them. (cadence.com)