OpenAI and Synopsys partner on GPT-Synopsys for AI-native chip design

Synopsys and OpenAI said Wednesday they signed a multi-year agreement to jointly develop GPT-Synopsys, a specialized model meant to operate Synopsys electronic design automation tools for semiconductor design and verification, with a shared revenue framework and early technology engagements already underway.

Chip design is how AI gets its hardware, and Synopsys just tied OpenAI to a specialized model that is supposed to drive the tools that draw those chips. What they announced is the partnership and the product plan: a model meant to work like an expert Synopsys user, with a customer’s design files kept off the training path, while early technology engagements are already underway.

On Wednesday, 30 September 2026, Synopsys and OpenAI announced a strategic partnership to jointly develop GPT-Synopsys. The newsroom post is dated that day and does not print an hour. PR Newswire carries the same release and stamps Sep 30, 2026, 14:29 ET, which is 2:29 p.m. Eastern. The dateline is San Francisco and Sunnyvale, Calif. Synopsys trades as Nasdaq: SNPS. The companies call GPT-Synopsys a specialized model, built to use Synopsys electronic design automation tools for semiconductor design workflows. Electronic design automation, shortened to EDA, is the software engineers use to design a chip and check it before a factory makes it. Those lines are the companies’.

The companies signed a multi-year agreement to collaborate as preferred partners. The highlights say they will develop and deliver GPT-Synopsys, a specialized model for chip design that brings OpenAI’s frontier AI together with Synopsys’ EDA tools and chip-design expertise. A frontier model, here, is a system at the top of what a lab says it can do. OpenAI will license Synopsys’ electronic design automation tools to develop the specialized model. The agreement includes a revenue-sharing arrangement and joint go-to-market work to make GPT-Synopsys available to customers worldwide. The highlights also say the two companies will work closely on research and on that joint selling effort, under a shared revenue framework. Those lines are the release’s.

The release describes what the partnership is supposed to change. Today, it says, agentic AI connects a general-purpose model to EDA tools and runs a chip-design workflow. Agentic, here, means software that takes a next step, such as running a tool, rather than only answering a question. The next step this deal describes is a frontier model that learns those tools the way an expert engineer would: running them, reading what they return, and using them to improve a design again and again. The release says that would make the model a native expert user, so engineers can try more design alternatives and optimize power, performance, and area. Power, performance, and area, shortened to PPA, is the three-way trade in a chip: how much electricity it uses, how fast it runs, and how much silicon it takes. Those lines are the release’s.

GPT-Synopsys, the release says, joins OpenAI’s frontier models with Synopsys’ EDA technology and the company’s knowledge of the field. The model is meant to reason about chip design and verification, and to operate Synopsys tools directly. Verification is the check that a design does what it was supposed to do. Engineers would hand the model objectives such as PPA optimization, timing, and verification closure. Timing is whether the signals inside the chip arrive when they must. Closure means that check is finished and the design meets the rule. Agents would run the tools, interpret the results, make changes, and iterate toward a verified outcome for an engineer to review. Those lines are the release’s.

GPT-Synopsys will run on computers OpenAI hosts. It is designed to work with the agent harness a customer already uses. A harness, here, is the software that aims an agent at a job and watches what it does. The model will be deeply integrated with Synopsys.ai, Synopsys’ own AI software, and with Synopsys Autopilot, which the release calls an agentic AI platform. Early technology engagements are underway with leading semiconductor customers. The release does not name those customers. Those lines are the release’s.

The joint service will bundle the compute, the model, and the licenses, and the release says customer-specific design data is protected. Compute, here, is the machines the model runs on. GPT-Synopsys will have enterprise-grade security, governance, and access controls. Customer data is not used to train the model. It is encrypted at rest and in transit. Encrypted at rest means the files are scrambled while they are stored. Encrypted in transit means they are scrambled while they move between machines. Retention, audit, and permission controls can be configured. Those lines are the release’s.

Sassine Ghazi, president and chief executive of Synopsys, said semiconductor engineering needs a much faster design process without giving up PPA or first-time-right silicon. First-time-right silicon means a chip that works the first time it comes back from the factory. He said the agreement will widen access to Synopsys’ advanced design capabilities and to the ground-truth engineering tools required to bring more complex chips to market. Ground truth, in that sentence, is the measured check a design has to pass. He said that with OpenAI, the companies are bringing frontier intelligence to chip design so more companies can develop advanced silicon, while keeping the rigor manufacturing requires. Greg Brockman, president and co-founder of OpenAI, said OpenAI is using its most advanced technology to improve the systems that power AI. He said that with Synopsys, that work comes to chip design, so engineers can explore more designs and reach a working chip faster, and that better chips can mean better AI for more people. Those quotations are in the release.

The release describes a partnership and a product plan. It says early technology engagements are underway. It does not give a calendar day for broad availability, and it does not give a price. A forward-looking note in the release says statements about development, capabilities, timing, availability, and expected benefits can turn out differently. On 23 September 2026, Synopsys announced a separate collaboration with TSMC on agentic design flows for TSMC’s A14 manufacturing process. That agreement is about design flows on a TSMC process. This one is with OpenAI, and the product it names is GPT-Synopsys.

Reuters reported the same announcement from San Francisco. Stephen Nellis’s story is dated 30 September 2026 and stamps 7:01 p.m. UTC, which is 3:01 p.m. Eastern. Ghazi told the wire that OpenAI will pay Synopsys a training subscription fee so the model can learn the tools, and that when customers use the product the two companies will share revenue based on how much the model improves a design. He said the agreement is built as added business, because it delivers more value to the customer. He also said the model’s work is still checked by Synopsys tools that use ordinary computing to test whether a chip will work. He said the model needs those guardrails to check the physics, and he called that check sign-off, or ground truth. Brockman, in a video the wire quotes, said the aim is to shave weeks or months off the design process and bring more chips into the world. The same story says the news came at an investor summit where Ghazi said Synopsys now expects 15 percent revenue growth in fiscal 2027, above an 11.19 percent analyst estimate from LSEG, and that the shares rose as much as 7 percent after the partnership and that outlook. Fifteen percent would be fifteen dollars of added revenue for each hundred dollars the year before. Seven percent is the largest move Reuters reports for the stock that day. Those lines are Reuters’s, and the analyst estimate is LSEG’s, as the wire cites it.

In plain terms, Synopsys and OpenAI said on Wednesday that they signed a multi-year deal to build GPT-Synopsys, a specialized model meant to run Synopsys’ chip-design tools. OpenAI will license those tools to develop the model. Engineers would set a goal, such as a better balance of power, speed, and size, or a finished timing and verification check, and agents would run the tools and hand a verified result back for a person to review. The model will run on OpenAI’s computers, work with a customer’s own agent setup, and sit inside Synopsys.ai and Synopsys Autopilot. Early work with leading chip customers has started. The release does not name them. Customer design data, the release says, is not used to train the model.

The picture is the official partnership graphic. A black field holds the Synopsys wordmark and the OpenAI blossom side by side. The headline reads “OpenAI and Synopsys Announce GPT-Synopsys,” with the line “Frontier Intelligence to Revolutionize Chip Design.” The footer prints San Francisco and Sunnyvale, Calif., September 30, 2026. It is the companies’ key art for this announcement. It is not a photograph of a chip.