Santa Clara, California, April 16, 2026, 09:30 PDT
Nvidia and Cadence Design Systems on Wednesday announced plans to pair Cadence’s physics engines—software used to model the behavior of materials and objects—with Nvidia’s AI models for training robots in simulated environments. The news came out of a Cadence-hosted event in Santa Clara and marks a further pivot from just making data-center chips, moving deeper into robotics.
Timing’s key here. Both TSMC and ASML boosted their forecasts this week, with TSMC CEO C.C. Wei calling AI demand “extremely robust.” Reflexivity’s Giuseppe Sette highlighted that ASML’s numbers offered a “favorable picture” for chips—despite all the usual AI-bubble jitters. Reuters
That leaves Nvidia with space to expand into fresh markets, even as its top buyers field intensifying questions about their returns. Reuters said Microsoft, Meta, Amazon, and Alphabet are on track to shell out over $600 billion on data centers this year. Nvidia, meanwhile, is shifting more of its attention to inference — that’s when trained models spit out results — and to software aimed at running physical-world machines.
Cadence described an expanded partnership reaching into agentic AI—software agents designed for multistep tasks—and digital twins, essentially virtual versions of hardware and data centers. According to Cadence, integrating its tools with Nvidia’s CUDA-X, Omniverse, and Isaac platforms could lift certain engineering workflows by as much as 100-fold, while also narrowing the gap between robots trained in simulation and those operating in real-world environments.
Nvidia CEO Jensen Huang, speaking at the event, told the audience, “We’re working with you across the board on robotic systems.” Cadence chief executive Anirudh Devgan weighed in, noting the impact of simulation data: “the more accurate (generated training data) is, the better the model will be.” Reuters
The deal throws more heat on Cadence. Back in March, Nvidia pointed out that Siemens and Synopsys, too, are developing AI agents for chip and system design using Nvidia hardware.
Nvidia’s dominance in robotics hardware is already established. Back in March, George Chowdhury at ABI Research told Reuters that more than 80% of humanoid robots run on Nvidia’s platform.
Robotics isn’t shaping up as a simple next step. “Nvidia is definitely going to see more competition compared to a year ago,” Summit Insights Group managing director KinNgai Chan told Reuters in March. Customers are pushing to develop their own inference chips; meanwhile, on Wednesday, Senator Elizabeth Warren pressed the Energy and Defense departments about whether Nvidia’s SchedMD deal could hand it excessive sway over software that’s vital for U.S. supercomputers. Reuters
Nvidia traded flat in early U.S. hours Thursday, holding at $198.95 and valuing the chipmaker around $4.53 trillion. Cadence, meanwhile, edged up roughly 1.1%, after a Reuters report noted the stock had jumped over 4% on Wednesday when news of the partnership broke.
Supply conditions remain tight at this point. “Demand for chips is outpacing supply,” ASML CEO Christophe Fouquet told reporters this week. Reuters
