AI agents are designing and testing chips: AgentEngineer release from Synopsys

Synopsys has recently announced the release of AgentEngineer. This new project is practically a set of specialized agents for the completion of long-term tasks that is based on the Autopilot AI platform. It covers chip verification, system-level verification, physical implementation, simulation, analysis, and analog/mixed-signal design.

Insights into the new product  

The new product is already implemented by more than 50 customers, and it is planned to be widely used by the end of 2026. The company describes the functioning of AI agents as autonomous, but human interference is still necessary for result validation. Such an approach can significantly improve efficiency by accelerating the transition from AI-assisted design to autonomous engineering.

The AgentEngineer system functions on the basis of a three-tier structure with superagent coordinators, executor agents, and tool-level engines that perform the requested work. They can process very challenging tasks with hundreds or thousands of reasoning steps.  

The platform covers major stages of development from work coordination to collection of telemetry. In addition, intellectual property security includes encryption, access control systems, and runtime environment protection mechanisms. Users can try open-source, commercial, and fine-tuned AI models within the new system, whereas the actual process can be conducted on on-premises infrastructure, in the client’s own cloud, or in the Synopsys cloud.

During joint system testing with Nvidia, the chip verification process sped up by almost 50 times, verification coverage increased by 20%, and productivity in solving engineering problems improved by 30%. The positive feedback about the system was already provided by Samsung, MediaTek, and Intel.  

Hope you found this article helpful - what do you think? Like and subscribe to our blog to get more practical tips and the latest tech news from HostZealot.