Synopsys and OpenAI Will Split the Revenue on an AI That Designs Chips
At its investor day on September 30, Synopsys agreed to co-develop an OpenAI model built to operate its own chip-design software, with the two companies sharing whatever revenue it generates. It's a bet that the moat in agentic AI is the tool, not the model — and it puts the question squarely to Cadence.
What Was Actually Agreed
On September 30, at its 2026 Investor Day, Synopsys and OpenAI announced GPT-Synopsys: a specialized AI model, trained to operate Synopsys's own electronic design automation (EDA) tools rather than replace them. OpenAI is licensing those tools to build the model on its own infrastructure; once it ships, the companies will jointly sell it and split the revenue based on how much it improves a customer's chip-design results. Reuters reported that OpenAI will also pay Synopsys a subscription fee to learn the tools in the first place. Neither the fee nor the revenue split has been disclosed.
Synopsys CEO Sassine Ghazi was explicit about why the deal is structured this way: "We structured the agreement in a way that it will not be cannibalizing our business," he told Reuters, framing it instead as "an upside to our business given we're delivering more value to the customer." Every output the model produces still gets checked by Synopsys's own verification tools before it's trusted — the model proposes, the existing software confirms.
Why Now
EDA has been a two-company market — Synopsys and Cadence — for decades, and both have spent 2026 racing to bolt agentic AI onto their tools. Cadence moved first in public, unveiling a Nvidia-powered "virtual engineer" at Computex in June that it says cuts verification cycles from five weeks to under a day. Synopsys's answer, announced four months later alongside a separate $1 billion-plus chip-design agreement with Amazon the same day, isn't just an in-house agent — it's a frontier lab's model, trained specifically on Synopsys's own tools and sold under a shared-revenue structure that ties OpenAI's upside to Synopsys's. Investors read it as validation: Synopsys shares jumped as much as 10% the next trading day.
What Each Side Gets
Neither the subscription fee nor the revenue-share formula was disclosed, so there's no way yet to say which side's bet is bigger.
Who It Affects
The direct target is Cadence, Synopsys's only real peer in EDA, which now has to decide whether to build its own frontier-model partnership — with Anthropic, Google, or another lab — or keep betting on in-house agents running on licensed models like Nvidia's Nemotron. Chip designers at Synopsys's customers (companies like Amazon, which signed its own $1 billion-plus silicon IP deal with Synopsys the same day) are the intended beneficiaries, assuming the model's claims hold up outside a press release. Nvidia sits in an odd spot: its chips train these models and its Nemotron models already power Cadence's rival agent, so it profits either way the EDA race shakes out.
What to Watch Next
Three things would turn this from an investor-day announcement into something real: whether Synopsys and OpenAI disclose an actual release date and revenue-split formula for GPT-Synopsys, rather than leaving both open-ended; whether Cadence's Q3 earnings call, expected around October 26, names a comparable frontier-lab partner of its own; and whether any customer — not Synopsys or OpenAI — reports a chip taped out meaningfully faster because of the model, the only claim in this announcement neither company can make on the other's behalf.
Sources
- Synopsys Investor Relations: OpenAI and Synopsys Announce GPT-Synopsys — primary announcement, September 30, 2026
- Reuters, via The Star: Synopsys, OpenAI strike deal to develop AI model for chip design work — October 1, 2026
- eWeek: OpenAI, Synopsys Build GPT-Synopsys to Let AI Agents Operate Chip Design Tools — October 4, 2026
- Business Wire, via Eagle-Tribune: Cadence Unveils Industry's First Fully Autonomous Virtual Engineer for Chip Design, powered by NVIDIA — June 1, 2026