Artificial intelligence requires increasingly powerful chips, and OpenAI now wants AI itself to play a bigger role in designing them. The company is partnering with Synopsys to develop GPT-Synopsys, a specialized model built to work directly with the tools engineers use to create semiconductors.
This is more interesting than simply putting a chatbot alongside chip design software. OpenAI and Synopsys say GPT-Synopsys will learn to operate electronic design automation, or EDA, tools much like an experienced engineer would. The model could run those tools, examine the results, make changes, and repeat the process as it works toward a particular design objective.
In other words, engineers could give GPT-Synopsys a goal involving power consumption, performance, chip area, timing, or verification. AI agents would then perform portions of the workflow using Synopsys software before presenting the results for human review.
The two companies have signed a multi-year agreement that goes well beyond a typical technology integration. OpenAI will license Synopsys EDA tools for development of the specialized model, while the companies will collaborate on research, sales, and bringing the resulting service to customers. The deal also includes a shared revenue arrangement.
There is something wonderfully circular about the whole idea. AI’s appetite for computing power has helped drive demand for increasingly sophisticated processors. Now OpenAI wants its models involved in designing the silicon that could eventually run future generations of AI.
Synopsys says the goal is to let engineers explore more potential designs while optimizing power, performance, and area. Chip development involves an enormous number of decisions and iterations, making it an obvious target for agents capable of repeatedly operating specialized engineering software.
GPT-Synopsys will run on OpenAI-hosted infrastructure and integrate with Synopsys.ai and the company’s Autopilot agentic AI platform. According to the companies, early technology engagements are already underway with semiconductor customers, although they have not identified those customers.
The handling of chip design data will understandably be important. Semiconductor designs can contain some of a company’s most valuable intellectual property. Synopsys and OpenAI say customer data will not be used to train the model and will be encrypted both at rest and in transit. Customers will also get controls covering retention, permissions, and auditing.
That doesn’t mean GPT-Synopsys is suddenly going to sit down and independently create the next flagship processor. The companies describe a system where engineers set objectives and review the resulting work. The more interesting development is that the AI isn’t limited to explaining what an engineer should do. It is being trained to actually use the engineering tools needed to do it.
If GPT-Synopsys works as intended, that could make AI agents considerably more useful inside semiconductor companies. Instead of serving as another assistant sitting beside an engineer, the model could take on repetitive iterations while humans remain responsible for the larger design and final decisions.
And yes, there is still plenty of marketing language surrounding the announcement. Synopsys says the partnership will “revolutionize” chip design, which is something nobody can know yet. What we do know is that OpenAI is moving its models deeper into a highly specialized engineering workflow where mistakes can become extremely expensive once a design reaches manufacturing.
The technology will ultimately have to prove that it can deliver useful designs without sacrificing the rigor chip development demands. But the basic idea is fascinating: AI helped create an enormous need for better chips, and now OpenAI wants AI to help design them.
Support independent tech journalism
NERDS.xyz is independently owned and operated. If you enjoy my coverage of Linux, AI, hardware, cybersecurity, and tech culture, consider supporting the site on Ko-fi.
Support NERDS.xyz
