日本語 
Cadence Design Systems and NVIDIA are not starting something new here. They are taking what already worked and scaling it harder across the AI and semiconductor stack.
The announcement came at CadenceLIVE 2026. The intent is simple if you strip the noise. Engineering workflows are too slow, too fragmented, and still depend heavily on manual steps. Both companies want to compress that cycle using agentic AI, simulation, and digital twins.
Cadence is plugging its EDA and system design tools into NVIDIA’s stack which includes CUDA-X, AI models, and Omniverse. The result they are pushing for is speed. In some cases, they are talking about up to 100 times improvement in certain engineering workflows. That is not a small claim, but it tells you where the pressure is coming from.
Also Read: Air Liquide’s $236M Investment Boosts Japan’s AI Chip Supply Chain
One big piece here is AgentStack. This builds on their earlier ChipStack AI agent and moves beyond just RTL design and verification. Now it stretches into physical design, analog systems, and full system-level workflows. Instead of engineers running scripts or clicking through tools, multiple AI agents coordinate tasks, make decisions, and push execution forward. NVIDIA is already using this internally, which basically makes them both the partner and the test case.
They are also going beyond chips. Physical AI is getting attention here. By combining Cadence’s simulation tools with NVIDIA’s robotics stack, they are trying to close the gap between simulation and real-world deployment. That gap has been a bottleneck for robotics and autonomous systems for years.
Then there is the AI factory angle. Both are working on digital twins for large-scale AI infrastructure. The focus is shifting to efficiency, especially cost per token. That means how much output you get for the power you consume. In one joint scenario, they showed meaningful gains in efficiency just by tuning power and cooling setups before anything is physically built.x`
Put all of this together and the direction is obvious. This is less about better tools and more about changing how engineering work actually happens. From manual and step-by-step to continuous, agent-driven execution.
English