The chip-software maker will share income with OpenAI from GPT-Synopsys, a model trained to handle design work, as Synopsys lifts its fiscal 2027 growth forecast above analyst expectations and sees its shares jump.
Synopsys, which makes software used to design computer chips, announced on Wednesday that it has agreed to share revenue with OpenAI as the two companies build an artificial intelligence model aimed at the semiconductor industry. The model, to be called GPT-Synopsys, will be trained and tuned specifically for chip design work.
The announcement came at an investor summit, where Synopsys chief executive Sassine Ghazi also said the company expects revenue to grow 15 per cent in its 2027 fiscal year. That is higher than the 11.19 per cent analysts had forecast, according to LSEG data. Shares climbed by as much as 7 per cent following the news.
What the model will do
Chip design begins with describing a processor's circuits in a code-like language and extends to deciding how billions of transistors should be arranged on a tiny piece of silicon. OpenAI's model will learn to operate Synopsys's software tools, helping engineers weigh the many trade-offs and optimisations involved at each stage.
OpenAI co-founder Greg Brockman said in a video announcing the partnership that the aim is to cut weeks or even months from the design process and allow more chips to reach the market.
How the money will flow
Under the arrangement, OpenAI will pay Synopsys a training subscription fee while its model learns to use the company's tools. Once Synopsys customers start using the product, the two firms will divide revenue according to how much the model improves a chip's design.
Ghazi said the structure was chosen so that it would not eat into Synopsys's existing business. He described it as an additional source of income, arguing that the company is delivering more value to customers.
Keeping humans and physics in the loop
Ghazi stressed that the AI's output will not go unchecked. Synopsys's conventional computing tools, which verify whether a chip will actually function, will review the model's work.
He said the model needs such safeguards to test the underlying physics, adding that validation at the highest level of accuracy, known in the industry as sign-off or ground truth, remains essential.
