OpenAI and Synopsys announced a multi-year partnership on 30 September to build GPT-Synopsys, a specialised model trained to drive Synopsys’ chip design software. The structure of the deal is the unusual part: the two companies will share revenue based on how much the model improves a customer’s chip design.
Synopsys sells electronic design automation software — the tools chip companies use to describe circuits, lay out transistors, close timing and verify that a design will work in silicon. Almost every advanced chip in production passed through software from Synopsys or one of its two main competitors.
What the model is supposed to do
Under the agreement, OpenAI licenses Synopsys’ EDA tools and trains a model to operate them. Engineers would hand the system an objective — a target for power, performance and area, or closing timing, or finishing verification — and agents would run the tools, read the results and iterate until the objective is met.
That is a narrower claim than it first sounds. The companies say traditional verification tools will still check the model’s output for physical correctness, so the model is being positioned as something that operates the existing flow faster rather than something that replaces the parts of the flow that guarantee a chip works.
The commercial structure
Neither company disclosed a price. The deal bundles compute, the model and Synopsys licences into a joint service, with joint go-to-market, and OpenAI says customer data is encrypted and not used to train the model.
Revenue sharing tied to design improvement is rare in enterprise software, where licences are normally sold by seat or by capacity. It implies both companies expect to be able to measure the improvement — which, in chip design, means agreeing with a customer on a baseline power, performance and area figure and then attributing the delta to the model.
Quotes and timing
Synopsys president and chief executive Sassine Ghazi framed the problem as speed without loss of quality: “The future of semiconductor engineering requires dramatic acceleration of the chip design process without compromising PPA or first-time-right silicon.” OpenAI president and co-founder Greg Brockman said the company is “bringing that work to chip design, helping engineers explore more designs and get to a working chip faster.”
Early technology engagements are already running with semiconductor customers. GPT-Synopsys will run on OpenAI-hosted infrastructure and integrate with Synopsys.ai and Synopsys Autopilot. Neither company gave a general availability date.
Why it matters beyond the two companies
OpenAI has spent the past year signing agreements that reach into the supply chain it depends on. A model that can drive EDA tools sits upstream of every accelerator the industry is waiting for, and the customers for it are the same firms OpenAI buys compute from.
What to watch
Whether any named customer publishes a measured result, and how the revenue share is actually calculated. Until a baseline methodology is public, the claim that the model improves a design is a claim the two parties to the contract are making about themselves.
