OpenAI and Synopsys are building GPT-Synopsys to do more than recommend chip-design changes: the specialized model is intended to operate Synopsys' electronic design automation tools itself. It would run tools, interpret results and iterate toward design goals before an engineer reviews the output.
The companies announced the multiyear partnership on Sept. 30, 2026, and said early technology engagements are underway with unnamed semiconductor customers. Pricing, a general availability date and customer benchmarks for production workflows have not been disclosed.
How GPT-Synopsys would operate chip design tools
Synopsys describes GPT-Synopsys as a planned “native expert user” of its EDA software. Engineers could assign objectives such as power, performance and area optimization, timing closure or verification closure, with agents implementing changes and iterating toward verified outcomes for engineer review, according to the Sept. 30 announcement.
OpenAI will license Synopsys EDA tools to develop the model. The companies also plan joint R&D and go-to-market work under a shared revenue framework, with a service bundling compute, model access and software licenses.
GPT-Synopsys is planned to run on OpenAI-hosted infrastructure and integrate with Synopsys.ai, Synopsys Autopilot and customer agent-harness systems. Synopsys says customer data will not be used to train the model and will be encrypted in transit and at rest, with configurable retention, audit and permission controls. The announcement does not detail the underlying isolation architecture.
The hosting model also puts agent containment and permissions on the procurement checklist. OpenAI recently tightened sandboxing, network access and permissions in its research systems after autonomous-model incidents, while Nvidia's Open Agent Safety Platform uses isolated runtimes and optional hardware monitoring to restrict agent behavior.
GPT-Synopsys extends Synopsys' existing agentic EDA work. In July, Synopsys introduced autonomous workflows developed with Microsoft and used by AMD; the company said early evaluations of a fully autonomous debug-closure workflow cut cycle time by 25% to 40%, according to its July agentic EDA announcement.
Other vendors are also bringing autonomous agents into chip-design workflows. eWeek previously covered Nvidia's $2 billion Synopsys deal, including plans to integrate AgentEngineer with Nvidia's agentic AI stack, while Siemens, Cadence and Synopsys also appeared in eWeek's Nvidia Agent Toolkit coverage of autonomous semiconductor workflows.
What eWeek Found: Agentic EDA has results, but GPT-Synopsys lacks public benchmarks
Synopsys has published measured results for autonomous EDA workflows, but not for GPT-Synopsys itself. Its July work with Microsoft reported a 25% to 40% reduction in debug cycle time in early evaluations, while the Sept. 30 GPT-Synopsys announcement gives no benchmark for timing closure, verification closure, PPA optimization or another defined customer workflow.
OpenAI offers a separate baseline. The company said AI helped move its Jalapeño inference chip from initial design to tapeout in nine months by shortening design, measurement and verification loops, according to OpenAI's August performance report. Earlier coverage of OpenAI's custom chip push shows how Jalapeño fits into its wider infrastructure strategy, but those results do not demonstrate GPT-Synopsys operating Synopsys tools inside an external customer's environment.
Semiconductor teams considering the service will still need workflow-level benchmarks, deployment and isolation details, retention and audit controls, human-approval boundaries, change provenance and pricing before putting proprietary designs into the system. A named customer result on a defined workflow would provide the clearest evidence of GPT-Synopsys' operational value.
Want to learn more AI tips, tricks, and prompting techniques? eWeek readers get free 7-day access to The Neuron Academy, our practical learning platform designed to help professionals use AI more confidently at work.
Let us teach you How to Talk to AI for free! Try our six-minute course at The Neuron Academy and learn a few simple ways to write better prompts and get more useful results from AI, or browse our other AI course for free for seven days. Check out all the lessons here →