GPT-Synopsys: OpenAI Teams With Synopsys on Chip Design AI

OpenAI has signed a multi-year strategic partnership with Synopsys to build a specialized AI model for chip design. The model is called GPT-Synopsys. The aim is to give semiconductor engineers a system that understands design work and can run the industry's design software directly.

The deal pairs a frontier AI lab with one of the established vendors in chip design software. Both companies are now testing whether a language model can take on part of the work of an experienced chip engineer.

Who Synopsys is

Synopsys sells electronic design automation (EDA) tools. Engineers use this software to design computer chips and other semiconductors. EDA covers much of the work between a chip concept and a manufacturable product, including design and verification, which checks that a design behaves as intended before it is built.

These tools are central to how modern chips are made. They are also complex, and using them well takes experience. That is where OpenAI thinks it can help.

What GPT-Synopsys is supposed to do

GPT-Synopsys combines OpenAI's AI technology with Synopsys' EDA tools. OpenAI is licensing those tools as part of the project. According to the companies, the goal is a model that can "reason about chip design and verification, and to directly operate Synopsys' tools."

That last part is the important one. The model is not meant only to answer questions about chip design or suggest snippets of code. It is meant to operate the software itself.

The planned workflow keeps humans in charge of decisions:

1. Engineers set the goal. They hand the model a design objective.
2. The model does the work. It reasons through the task and uses Synopsys' tools to carry it out.
3. Engineers review and approve. People check the output and decide whether to accept it.

This is the agent pattern that has spread across software development over the past year, now applied to a narrow and costly engineering field. Delegation with an approval step is a familiar design choice. It gives the model room to work while leaving consequential decisions with people.

Data, infrastructure and business terms

GPT-Synopsys will run on OpenAI's infrastructure. Chip designs are among the most closely guarded assets a semiconductor company has, so data handling will matter to any potential customer. Both companies say customer data will not be used for training and will be stored encrypted.

Early tests with semiconductor customers are already underway. The companies have not named those customers in the source material.

The commercial setup is a joint one. OpenAI and Synopsys will market GPT-Synopsys together and share the revenue.

What the executives are saying

Synopsys CEO Sassine Ghazi says AI could significantly speed up the chip design process. OpenAI co-founder Greg Brockman describes the partnership as a route to both better chips and better AI.

Brockman's framing points to a loop OpenAI clearly cares about. Better AI could help design better chips, and better chips could then run better AI. The company already has a direct stake in that second half.

OpenAI's wider chip ambitions

OpenAI is separately working with Broadcom on chips built specifically for running AI models. The recently unveiled Jalapeno chip came out of that work and, according to the report, is very competitive against similar specialized chips.

So OpenAI is now involved in chips from two sides. With Broadcom, it is a customer and co-developer of hardware for its own workloads. With Synopsys, it wants to become a supplier of AI that helps other companies design hardware.

Our Take

This deal suggests that AI labs see chip design as a serious commercial target rather than a side experiment. EDA is a specialist market with a small number of major vendors, and partnering with one of them gives OpenAI access to tools and customers it could not easily reach alone. For Synopsys, the partnership looks like a way to put a frontier model inside its products instead of competing with one.

It also fits a broader pattern. AI companies are moving closer to the hardware they depend on, from OpenAI's custom silicon work with Broadcom to Deepseek releasing open-source tools for Huawei Ascend chips. Large players are also making expensive bets on adjacent technology, such as AMD's purchase of World Labs. Compute is a constraint for every lab, and owning more of the stack is one response.

There are open questions. The source does not include any performance figures, so claims about faster design cycles remain promises for now. It is worth watching whether early customers report measurable time savings, and whether the human approval step holds up in practice. Review steps work only if engineers actually check the output instead of approving by habit.

Trust is the other issue. Chipmakers will need to be comfortable sending sensitive design work to models running on OpenAI's infrastructure, even with encryption and a no-training commitment. How quickly customers move from tests to production use will show whether those assurances are enough.

Finally, it is worth watching how competing EDA vendors and other AI labs respond. If GPT-Synopsys proves useful, similar partnerships elsewhere in the chip industry seem likely.