Synopsys’s headquarters building at 675 Almanor Avenue in Sunnyvale, California, in 2024 (file photo). Image: Coolcaesar / Wikimedia Commons, CC BY 4.0, cropped
OpenAI wants its AI to help design computer chips, including the kind that power AI itself. The company and chip design software maker Synopsys announced on Wednesday a multi-year partnership to build GPT-Synopsys, a specialised model trained to use Synopsys’s design tools the way an expert engineer would.
What GPT-Synopsys is meant to do #
Almost every advanced chip is designed with electronic design automation (EDA) software, and Synopsys is one of the biggest makers of it. Engineers use these tools to lay out billions of transistors, check timing, and squeeze out the best balance of power use, performance and chip area, which the industry calls PPA.
AI agents can already drive EDA tools by connecting a general-purpose model to them. The companies say GPT-Synopsys goes a step further: the model is being trained to run the tools itself, read their results and keep improving a design, becoming what the announcement calls “a native expert user”.
In practice, engineers would hand the model a goal, such as hitting a power target or closing timing and verification, and agents would run the tools, interpret the output, make changes and iterate “toward verified outcomes for engineer review”. The engineer still signs off.
How the deal works #
- OpenAI licenses Synopsys’s EDA tools to train the model, and the two companies will work together on research.
- Revenue is shared. They will sell GPT-Synopsys together to customers worldwide as a bundle of compute, model and software licences.
- It runs on OpenAI’s servers , plugs into Synopsys’s own agent platform, Synopsys Autopilot, and is designed to work with the agent systems chipmakers already use.
- Customer data isn’t used for training , according to the companies, and is encrypted, with controls over retention, auditing and permissions.
The companies say early work is already under way with “leading semiconductor customers”, but didn’t name them, give prices or say when GPT-Synopsys will be generally available. The performance claims are the companies’ own and haven’t been independently tested.
“We can build better AI” #
OpenAI president Greg Brockman framed the deal as AI improving the hardware it runs on:
We’re using our most advanced technology to improve the systems that power AI. With Synopsys, we’re bringing that work to chip design, helping engineers explore more designs and get to a working chip faster. By helping them build better chips, we can build better AI and bring it to more people.
Greg Brockman, president and co-founder, OpenAI
Synopsys chief executive Sassine Ghazi said chip design needs to speed up dramatically “without compromising PPA or first-time-right silicon”, the industry’s term for a chip that works on its first manufacturing run, when mistakes can cost millions.
That loop, AI helping to design the chips that train the next AI, is exactly what makes the deal notable. Demand for AI chips has turned the industry upside down this year, with AMD passing a $1 trillion valuation and rivals such as DeepSeek now writing their own chip software to escape Nvidia. It also comes a day after OpenAI’s DevDay, where it pushed agents that work on their own for hours.
Why it matters #
Designing a cutting-edge chip takes large teams years, and any tool that shortens that cycle speeds up the whole AI industry. If GPT-Synopsys works as promised, AI will be helping to build its own hardware, which could make better chips cheaper and faster to design, and gives OpenAI a foothold in one of tech’s most specialised trades.