Cadence launches Maestro AI models for in-house chip design

Cadence Design Systems introduced Maestro, a portfolio of domain-specialized AI models designed to support chip and system-design workflows. The company said customers will be able to run the models through Cadence's cloud, a public cloud or their own data centers and firewalled environments.

The launch places Cadence in the growing race to make AI agents useful inside electronic-design-automation software. Chip design involves arranging billions of transistors and meeting performance, cost and other engineering constraints. Cadence and rival Synopsys are developing different approaches to bringing models into those workflows.

Model choice and deployment control

Cadence said Maestro combines domain-specialized models, trained decision models and agent skills with its EDA and system-design tools. Customers can also fine-tune the models on proprietary designs and workflows, according to the company's announcement. The portfolio is intended to work alongside frontier and open-weight models rather than forcing customers into a single model source.

NVIDIA contributed technology to the collaboration. Reuters reported that Maestro was built in part on NVIDIA's open-weight Nemotron model and uses NVIDIA's OpenShell safety technology. Cadence's announcement described the partnership as combining Nemotron, OpenShell and engineering expertise, with reinforcement-learning feedback from EDA tools and deterministic solvers used to evaluate outputs.

Performance claims await broader testing

Cadence said initial design-workflow results showed up to 30% higher quality and up to five times greater token efficiency than a general-purpose foundation model. These are company-reported early results, not an independent benchmark presented in the Reuters report or Cadence announcement. They should therefore be read as performance claims tied to Cadence's test environment rather than a guaranteed result for every customer or design.

The company said its first Maestro models would reach select early-access customers in the fourth quarter of 2026. That timing describes an early-access rollout, not general availability or proof of production-scale adoption.

Competition in design software

Cadence's model-agnostic positioning contrasts with Synopsys' recently announced plan to develop an AI model with OpenAI and share revenue with the ChatGPT maker. Cadence senior vice president Paul Cunningham told Reuters that both open- and closed-weight approaches were being pursued and that work with commercial AI laboratories remained part of the Maestro vision.

The immediate significance of Maestro is control: chip companies can combine specialized models with existing design tools while keeping deployment in a cloud or on-premises setting that fits their data and security requirements.

Sources

• Reuters: Cadence unveils NVIDIA-backed AI models that chipmakers can run in-house
• Cadence: Cadence introduces Maestro AI Model Portfolio