Updated Version 1 - Cadence launches a set of AI models that chip manufacturers can run internally

Content rewritten: incorporates a broader suite of models including those supported by Nvidia, adds stock price information, and revises paragraphs 1 and 5.

Stephen Nellis

Reuters San Francisco, October 26 - Cadence Design Systems (CDNS.O) on Monday launched a suite of AI models optimized to assist in computer chip design, deepening its cooperation with Nvidia (NVDA.O) in the open-source model field, and distinguishing itself from rival Synopsys’s recent deal with OpenAI (link).

Both companies develop software for designing computing chips—a process that involves precisely arranging billions of transistors to meet cost and performance goals—and are racing to allow AI “agents” to use these tools on behalf of human engineers to accelerate the design process.

The companies stated that the new suite introduced Monday, called “Cadence Maestro,” is partly built on Nvidia’s open-weight Nemotron model and utilizes Nvidia’s OpenShell security technology (link) for protection.

Cadence said customers can run these models on their own cloud or any public cloud, or keep them entirely within private firewalls and data centers.

Cadence will release earnings after the close on Monday; following this announcement, its stock price rose by 3.9%.

The company added that customers can also fine-tune the models for their own chip designs, with early results showing up to a 30% improvement in quality and up to 5 times greater cost efficiency compared with general-purpose models.

Cadence’s move stands in sharp contrast to its largest competitor Synopsys. On September 30, Synopsys announced a partnership with OpenAI to develop an AI model and share revenue with the ChatGPT maker. This model, called GPT-Synopsys, will run on infrastructure hosted by OpenAI.

Paul Cunningham, Cadence senior vice president, positioned Maestro as a “model-agnostic” solution, saying: “There are many more (AI) model providers than (chip design software) vendors.”

Asked whether Cadence would sign deals with proprietary AI labs, Cunningham said the company is giving “equal weight” to both open and closed approaches, that partnering with commercial labs is “part of the Maestro vision,” but declined to elaborate further.

Cadence said the first Maestro models will be available to select early-access customers in the fourth quarter.

(To facilitate non-native English speakers, Reuters has automated translations of its reports into several other languages. As automated translations may contain errors or lack necessary context, Reuters does not guarantee the accuracy of automated translated texts and provides them purely for reader convenience. Reuters bears no responsibility for any damage or loss incurred as a result of using the automated translation feature.)