Synopsys has unveiled AgentEngineer, an agentic AI platform that autonomously plans and executes semiconductor engineering tasks, and says deployments at more than 50 customer projects — including Samsung Electronics, Nvidia and Intel — have already delivered up to 50x faster verification closure.

The company reported additional results from those customer projects: up to 30% productivity gains and up to 20% improvement in verification coverage. Token efficiency for AI model usage improved by up to 2x. Synopsys has scheduled the official launch of the platform for year-end.

Alongside AgentEngineer, Synopsys introduced Autopilot, an integrated operations foundation that orchestrates the agents. The announcement, made on the 2nd, positions the EDA vendor at the front of a fast-forming agentic AI race in chip design tools — AMD unveiled its own agentic assistant, "AMD Ross," just one day earlier.

From Commands to Goals

The architectural shift is the point. Existing AI assistants relied on engineers inputting individual commands. AgentEngineer instead receives only a goal, then autonomously decomposes the required tasks, formulates a plan, and draws on Synopsys design and verification tools to produce results. AI agents handle reasoning, planning and execution across engineering disciplines spanning silicon design to system development.

Application areas cover semiconductor design verification, physical design, analog design, semiconductor manufacturing, and structural, fluid and electromagnetic simulation. Synopsys operates domain-specific AI agents alongside task-level agents that handle individual work items.

In verification, agents perform coverage closure — confirming a design has been sufficiently tested against target behavior. In physical design, the platform automates the work of driving performance, power and area (PPA) to target levels. In manufacturing, capabilities include mask synthesis.

Open Architecture and IP Protection

Autopilot coordinates which tasks agents perform, retains knowledge and prior work context during the design process, and manages execution workflows and access permissions.

Synopsys adopted an open architecture that does not lock customers into its own AI models and tools. Customers can select models, computing infrastructure, data and agents from Synopsys, its partners, and external vendors.

Because semiconductor development involves sensitive design data and intellectual property, the platform incorporates access controls, encryption and execution-stage safeguards. AI agents operate only within boundaries the enterprise sets when accessing external data or tools.

Synopsys envisions moving beyond AI-assisted engineering to what it calls "autonomous engineering," where AI continuously uses multiple design tools to handle the entire development process.

Choi Jung-yeon, vice president of memory design technology at Samsung Electronics, said: "Synopsys' agentic AI solutions enable Samsung engineering teams to accelerate complex development workflows, allowing them to focus more on advanced memory technology innovation that will power next-generation AI infrastructure."

Ravi Subramanian, chief product management officer at Synopsys, said: "Customers are redesigning engineering workflows across semiconductor and system products to address rising system complexity and aggressive time-to-market schedules. We will accelerate the transition from AI-assisted design to autonomous engineering."

Competition Arrives Fast

The timing signals how quickly agentic AI competition has intensified across the design tools market. AMD Ross supports embedded system development across hardware and silicon design, board design, verification and debugging, and AMD has begun offering developer support.

AMD Ross interprets conversational instructions from developers — document search, command execution, debugging, design optimization — and connects directly to development tools to execute the work. It is not tied to any specific large language model or IDE; it links AMD's embedded development tools with AI agents through Model Context Protocol (MCP) servers, an open standard. It also supports air-gapped environments disconnected from the internet.

Cadence has likewise introduced an agentic AI design platform that goes beyond AI-based design space exploration and PPA optimization to simultaneously optimize multiple blocks. Samsung Electronics is pursuing circuit design automation using reinforcement learning and generative algorithms, along with multi-agent workflows in which AI agents exchange results across design stages.

Industry observers expect AI's role to expand as semiconductor design grows more complex and development cycles shrink. Particular attention falls on automating verification — the most time-consuming stage before commercialization — where Synopsys' reported 50x closure improvement directly addresses the bottleneck.

Security and access control remain the critical challenges, since agentic AI handles design data and IP. Synopsys' emphasis on access controls, encryption and execution-stage safeguards, and AMD's air-gap support, both reflect the requirements of development environments where IP protection is paramount. Whether Synopsys' early deployment numbers hold across broader production use after the year-end launch will shape how quickly autonomous engineering displaces AI-assisted design across the industry.