OpenAI and Synopsys partner on AI system for chip design

OpenAI and Synopsys have announced a strategic, multi-year partnership to develop GPT-Synopsys, an AI model built to work directly with Synopsys’ electronic design automation (EDA) tools. According to EEJournal, the agreement also covers revenue sharing and a joint effort to bring the service to customers around the world.

The project combines OpenAI’s frontier models with Synopsys’ chip-design software and engineering expertise. Its aim is to move beyond using general-purpose AI as a simple link to design tools. Instead, GPT-Synopsys is intended to learn how experienced engineers use those tools, understand the results they produce and repeatedly refine designs based on that analysis.

That approach could give chip designers more opportunities to test alternatives while working to improve power, performance and area, commonly referred to as PPA. The companies say the system is intended to help teams develop increasingly complex semiconductors more quickly without abandoning the checks needed to produce working silicon on the first manufacturing attempt.

From AI assistance to tool operation

Current agentic AI systems can connect general-purpose models with EDA software so that automated agents carry out parts of a chip-design workflow. The proposed GPT-Synopsys model would be more specialised: it would reason about semiconductor design and verification while directly operating Synopsys tools.

Engineers would set objectives such as improving PPA, completing timing work or reaching verification closure. Agents could then run the relevant tools, assess their outputs, make design changes and continue the process until they reach results that engineers can review. The model is not described as replacing that review; rather, it is designed to handle more of the iterative work that can consume engineering time.

The companies present this capability as a way to widen the search for viable designs. Chip development involves balancing several competing requirements, and exploring additional possibilities can be difficult when teams must perform each step manually. A system able to execute and evaluate repeated design iterations could help engineers investigate more options within the same development process.

A bundled service for semiconductor customers

GPT-Synopsys will operate on infrastructure hosted by OpenAI. It is also planned to work with customers’ agent-harness systems and to connect closely with Synopsys.ai and Synopsys Autopilot, the company’s agentic AI platform.

The offering is expected to combine computing resources, the specialised model and the required software licences. The companies say it will include enterprise security, governance and controls over access. Customer design information will not be used to train the model, according to the announcement, and data will be encrypted both while stored and while moving between systems.

Customers will also have configurable controls for data retention, auditing and permissions. These measures are intended to address the sensitivity of semiconductor designs, which can contain valuable intellectual property and details of products that have not yet reached the market.

Early technology engagements with leading semiconductor customers are already under way. The announcement does not identify those companies or provide a timetable for broader availability.

Why the partnership matters

The collaboration links a provider of advanced AI models with a company whose tools are used across semiconductor design, verification and analysis. Synopsys describes itself as an engineering-solutions provider spanning silicon to systems, with products covering chip design, intellectual property, simulation, analysis and design services.

For OpenAI, the partnership applies its AI technology to the creation of the hardware needed to run increasingly capable computing systems. Greg Brockman, OpenAI’s president and co-founder, said the work is intended to help engineers build better chips, which could in turn support better AI systems and broaden access to them.

Synopsys chief executive Sassine Ghazi said the agreement is designed to accelerate chip development while preserving the power, performance and area targets, as well as the discipline required for manufacturing. The companies’ commercial arrangement means the project is not only a technical integration: they will also jointly take the resulting service to the semiconductor market.

By embedding a frontier model inside established EDA workflows, the partners are targeting a specific part of the technology supply chain rather than offering a general design assistant. The success of the effort will depend on how effectively GPT-Synopsys can interpret engineering results, make useful changes and fit within the controls and review processes used by chip teams.