OpenAI and Synopsys are building GPT-Synopsys, an AI model that runs chip-design software itself

The agentic AI wave has reached the most expensive software in engineering.

On September 30, Synopsys and OpenAI announced a multi-year strategic partnership to build GPT-Synopsys, a specialized model trained to operate Synopsys's electronic design automation tools directly. OpenAI will license Synopsys's EDA software to train the model, and the two companies will jointly sell the result through a bundled offering of compute, model access, and licenses, with a revenue-sharing arrangement between them.

From assistant to expert tool user

The distinction from today's AI-assisted chip design is the point. Right now, general-purpose models connect to EDA tools through agentic wrappers to run individual workflows. Synopsys says GPT-Synopsys is meant to become an expert user of the tools themselves: learning to run them the way experienced engineers do, interpreting their outputs, and iteratively optimizing a design.

Engineers would delegate objectives like power, performance, and area (PPA) optimization or timing and verification closure, and the model would invoke the tools, examine violations and tradeoffs, change the design, rerun the workflow, and iterate until it has something worth an engineer's review.

"The future of semiconductor engineering requires dramatic acceleration of the chip design process without compromising PPA or first-time-right silicon," said Synopsys president and CEO Sassine Ghazi. "Together with OpenAI, we're bringing frontier intelligence to chip design to help more companies develop and accelerate advanced silicon."

Revenue tied to results

The commercial structure is unusual and worth noting. Reuters reported that OpenAI will pay Synopsys a training subscription fee while the model learns the tools, and that once customers deploy the product, the companies will share revenue based on how well the model actually improves chip designs. OpenAI co-founder Greg Brockman said the goal is to "shave off weeks, months from the design process and to bring more chips to the world."

Ghazi told Reuters the deal was structured so it won't cannibalize Synopsys's core business: "We structured the agreement in a way that it will not be cannibalizing our business. It will be an upside to our business given we're delivering more value to the customer."

There are guardrails baked into the pitch. Synopsys says customer-specific design data won't be used to train the model, that data is encrypted in transit and at rest with configurable retention, audit, and permission controls, and that anything the model produces still gets verified by Synopsys's traditional sign-off tools. "The model needs these guardrails in order to check the physics," Ghazi said. "The need for validating with the highest level of fidelity, what we call sign-off or ground truth, is essential."

What ships, and when

The model will run on OpenAI-hosted infrastructure, integrate with Synopsys.ai and the company's Autopilot agentic platform, and is designed to interoperate with customers' own agent harness systems. Synopsys says early technology engagements are already underway with leading semiconductor customers.

The announcement landed alongside Synopsys's investor day, where the company also forecast 15% revenue growth for fiscal 2027, above analysts' 11.19% estimate cited by Reuters, and introduced AgentEngineer, a set of domain-specific long-horizon agents built on the Autopilot platform with availability planned for the end of 2026. Shares rose as much as 7% after the announcements.

For developers, the signal is where the agents are going: not just writing code, but learning to drive the specialized software that other engineers spend careers mastering. Chip design is the extreme case. It won't be the last.