Synopsys and Ansys power the future of innovation—connecting silicon to systems.
EDA
System
IP
SolvNetPlus
Training & Education
No two chip design teams work the same way.
Every semiconductor company has built its own flows, scripts, and methodologies, often tuned to a specific product line. Those differences, sometimes shaped over decades, reflect each company’s unique engineering DNA and help define its competitive advantage.
So when a design team asks how to adopt agentic AI, the most useful answer starts with two questions: What do you already have, and what do you want to own?
Some teams want a complete solution they can put to work right away. Others have invested time and effort developing their own agentic platforms and want to build on those investments. And some may want to bring frontier intelligence to their most complex engineering challenges.
There is no single path that works for every company. That’s why we built our agentic AI portfolio around customer choice.
Why agentic AI, and why now
Chip complexity is rising and development schedules are compressing. Designs have moved from single dies to multi-die systems with hundreds of billions of transistors. Engineering teams are already under pressure to accomplish more with limited resources. And more companies are designing their own silicon, increasing demand for those same resources.
Agentic AI can help relieve these constraints.
The right adoption path, however, depends on each company’s data security requirements, compute environment, internal tooling, and the parts of the workflow it wants to control.
Agentic AI built for engineering
Our agentic portfolio is built on Synopsys AgentEngineer™ solutions and the Synopsys Autopilot™ Platform. AgentEngineer solutions are long-horizon super agents designed to take on complex engineering objectives. They can plan the work, enlist specialized task agents, use Synopsys engineering tools, evaluate results, and adapt their approach until the objective is achieved.
We have introduced AgentEngineer solutions across verification, implementation, analog and mixed-signal design, manufacturing, and simulation and analysis.
The Autopilot Platform provides the context, persistent memory, governance, orchestration, and tool connections those agents depend on, and we designed it to be open. Customers can integrate their own agents, connect with their existing engineering environments, and access Synopsys knowledge graphs, proprietary APIs, and tool data. Open protocols such as MCP and A2A help connect Synopsys agentic capabilities with customers’ existing environments, while giving them flexibility in the models they use.
The AI layer can vary. The engineering ground truth cannot.
Engineering teams may choose different platforms, agents, and models, but the resulting designs still need to be generated and verified using deterministic, physics-based engineering tools. With that foundation, customers have multiple ways to adopt agentic AI.
Path one: The full Synopsys stack
Some customers want the complete Synopsys solution: the Autopilot Platform, AgentEngineer solutions and specialized task agents, and Synopsys EDA tools working together.
Engineers define the objective, constraints, and success criteria, and the integrated environment orchestrates the work across agents and tools.
Because the agents have native access to Synopsys tools through proprietary APIs, they can act on results, evaluate progress, and adjust their approach without engineers having to connect and manage each step themselves. This gives customers a ready-to-use, full-stack agentic AI solution without requiring them to build the underlying platform or integrations.
Path two: A customer-built or hybrid stack
Many companies have invested heavily in their own agentic platforms and treat those workflows as a competitive advantage. They don’t want to replace them.
At the same time, they may want to license a proven capability their stack doesn’t include rather than build it from scratch. In this path, customers preserve the platforms, models, and proprietary agents they have built while connecting them with Synopsys agentic capabilities and trusted engineering tools through the open Autopilot architecture.
Customers decide which agents direct each workflow.
Path three: A frontier-model-led stack
The third path pairs a frontier model with Synopsys tools, agents, workflows, and engineering knowledge. GPT-Synopsys, developed with OpenAI and post-trained for chip design, is one example.
OpenAI contributes frontier reasoning, while Synopsys provides the domain expertise and ground-truth engineering capabilities. The Synopsys knowledge used to post-train the model remains proprietary within GPT-Synopsys rather than flowing back into the base model.
Customers access it as a service, and early engagements are already underway.
The common ground across every path
Whichever path a company takes, the design still has to pass signoff. AI can reason through options and propose solutions, but tape-out still depends on deterministic, foundry-certified tools to verify a design is ready for manufacturing.
Customers decide who provides the platform, agents, and model. Synopsys ground-truth EDA and simulation engines generate and validate the results. As our customers have told us: In AI we believe, but in physics we trust.
This flexibility in the AI layer, backed by trusted engineering tools, allows customers to adopt agentic AI on their own terms and move from an objective to a verified design — faster and with greater confidence.
