OpenAI and Synopsys build GPT-Synopsys for chip design

OpenAI and Synopsys have signed a multiyear deal to build GPT-Synopsys, an AI model trained to operate Synopsys's chip-design software. The companies announced it on September 30, 2026, and Synopsys stock closed 4.77% higher that day, according to StockTitan. The deal puts a frontier AI model inside one of the most expensive and specialized jobs in tech: designing a chip.

What GPT-Synopsys is

Synopsys makes electronic design automation (EDA) software. These are the programs engineers use to lay out a chip and check that it works before it is manufactured. GPT-Synopsys is a specialized model meant to operate those tools and run chip-design workflows, Converge Digest reports.

Engineering.com lists three kinds of work it will support:

• PPA optimization. PPA stands for power, performance and area, the three numbers chip designers trade against each other.
• Timing closure. This is the step that makes sure every signal arrives in time at the chip's target speed.
• Verification. This proves the design does what its specification says.

Early technology work with semiconductor customers is already underway, all three sources say. Neither company has given a price or a date for general availability, Converge Digest notes.

How the deal works

The model will not run on customers' own machines. It runs on servers hosted by OpenAI and is sold as one service that bundles computing power, the model and Synopsys software licenses, StockTitan's copy of the announcement says.

Part of the deal	What was announced
Length	Multiyear agreement
Status	Each company names the other a preferred partner
Model training	OpenAI licenses Synopsys EDA tools to develop the model
Business terms	Joint research, joint sales and revenue sharing
Hosting	OpenAI-hosted infrastructure
Product links	Works with Synopsys.ai and Synopsys Autopilot
Customer data	Not used to train the model; encrypted at rest and in transit
Price and launch date	Not disclosed

Both Synopsys.ai and Autopilot are existing Synopsys products, so the model is meant to slot into tools many chip teams already use.

What the companies say

Synopsys CEO Sassine Ghazi framed it as a speed play. "Together with OpenAI, we're bringing frontier intelligence to chip design to help more companies develop and accelerate advanced silicon," he said, as quoted by StockTitan. Converge Digest quotes him saying chip design needs "dramatic acceleration" without hurting PPA. He also wants chips that work on the first manufacturing run, which the industry calls first-time-right silicon.

OpenAI President Greg Brockman described a loop. "By helping them build better chips, we can build better AI and bring it to more people," he said. OpenAI has a direct interest in that loop: it already designs its own inference chip, Jalapeño, with Broadcom.

How investors reacted

Investors read the deal as good news for Synopsys. Its shares closed 4.77% higher on September 30, StockTitan reports. Trading volume reached 3.4 times the daily average, and the company's market value stood at $85.39 billion.

What this means for developers

For chip teams, nothing ships yet. Access is limited to early engagements, with no price and no release date. If you use Synopsys tools, ask your account team how to join and what a trial would cover.

Read the data terms first. Your design files would be processed on OpenAI-hosted servers, not inside your own network. The companies say customer data will not train the model and is encrypted at rest and in transit. Many chip companies still treat designs as their most sensitive asset, so legal and security teams will want to review the setup before any real project goes in.

Start any trial on a low-risk block of a design. Compare the model's results on power, timing and area against what your engineers produce today. Keep human sign-off on verification, because a missed bug in silicon cannot be patched like software.

For software developers, the pattern matters more than the product. GPT-Synopsys is a frontier model trained to drive one company's professional tools, sold with that company's licenses and a revenue share. Expect more deals shaped like this in other expert software. When they arrive, the questions are the same: who hosts it, who sees your data, and how the tool licenses get priced.