OpenAI wants to design the chips that one day might run OpenAI models. On September 30, 2026, the company announced a multi-year strategic partnership with Synopsys, the dominant maker of electronic design automation (EDA) software, to build a specialized AI model called GPT-Synopsys that’s meant to operate chip-design tools the way an experienced engineer would.
Here’s what the partnership actually involves, and why it matters for both companies.
What GPT-Synopsys is supposed to do
Chip design starts with a description of a circuit written in a code-like hardware description language, and ends with figuring out how to physically lay out billions of transistors on a small square of silicon — a process full of trade-offs between performance, power, and area that engineers currently navigate largely by hand, aided by existing EDA software. GPT-Synopsys is built to reason about chip design and verification and to directly operate Synopsys’ own tools, rather than simply assisting an engineer who’s still driving every step. OpenAI co-founder Greg Brockman described the goal plainly: shaving weeks or months off the design process so more chips can reach the world faster.
Importantly, this isn’t meant to replace human engineers entirely. Engineers will delegate design objectives to the model and then review and approve its output — and Synopsys CEO Sassine Ghazi confirmed that the AI’s work will still be double-checked using Synopsys’ traditional, non-AI verification techniques to confirm a chip will actually function correctly before it goes anywhere near manufacturing. In an industry where a design flaw can mean a multi-million-dollar failed production run, that verification safety net isn’t optional — it’s the whole reason Synopsys’ tools are trusted in the first place.
The business structure: shared risk, shared reward
Rather than simply licensing its technology to Synopsys or building a one-off tool, OpenAI and Synopsys structured this as a genuine revenue-sharing partnership. When Synopsys customers use GPT-Synopsys, the two companies will share revenue based specifically on how well the model actually improves a chip’s design — tying OpenAI’s payout directly to real, measurable engineering outcomes rather than simple usage or subscription volume. The deal also includes a joint global go-to-market strategy, meaning the two companies will sell this together rather than Synopsys simply bolting a third-party AI feature onto its existing product line.
The market reacted immediately: Synopsys shares rose as much as 7% following the announcement, which came alongside the company raising its fiscal 2027 revenue growth expectations at an investor summit.
Why Synopsys needed this deal
The partnership reads as a defensive move as much as an offensive one. Synopsys has spent the last two years watching generative AI creep into territory its EDA software has historically owned — and rather than risk an outside AI company eroding its core franchise by building a competing tool from scratch, Synopsys chose to build the model together with the AI developer itself and share the proceeds. It’s a hedge: if AI is going to fundamentally change how chips get designed regardless, better to be the company whose tools and verification expertise sit underneath that AI than to be disrupted by it.
Synopsys wasn’t starting from zero on this front, either. The company’s own fiscal 2025 results already pointed to this direction — the company highlighted GenAI capabilities that accelerated design workflows from days to hours, and hours to minutes, alongside an early technology framework called AgentEngineer, aimed at the same agentic chip-design frontier GPT-Synopsys now formalizes.
Why OpenAI wants this, beyond the obvious revenue angle
There’s a more self-interested read on why OpenAI is pursuing this partnership too. OpenAI is already working with Broadcom on chips purpose-built for running AI models — including a recently unveiled chip called Jalapeno, which is reportedly highly competitive against similar specialized AI accelerator chips from other companies. A deeper, more capable AI tool for chip design doesn’t just generate new revenue for OpenAI; it potentially feeds directly back into OpenAI’s own hardware ambitions, giving the company a more sophisticated internal capability for designing the specialized silicon its own AI systems increasingly depend on.
Already in testing
This isn’t purely a future roadmap item — early tests with real semiconductor customers are already underway, according to reporting on the announcement, suggesting GPT-Synopsys is further along than a typical “we’re exploring a partnership” press release would imply.
The bigger picture
This deal is a clear signal of where the AI industry’s next major expansion phase is heading: out of chatbots and coding assistants and into highly specialized, technically demanding industrial workflows where the margin for error is tiny and the value of getting it right is enormous. Chip design — with its multi-million-dollar manufacturing runs, multi-month timelines, and brutal transistor-level trade-offs — is about as high-stakes a proving ground as AI could choose. If GPT-Synopsys genuinely delivers faster first-pass silicon without compromising reliability, it won’t just be a win for two companies’ stock prices — it could meaningfully compress how quickly new, more capable chips (including, eventually, the ones training the next generation of AI models) actually reach the market.
Leave a Reply