Overview
- Cadence publicly unveiled Maestro on Monday as a portfolio of domain-specific AI models aimed at automating chip design tasks and speeding engineering workflows.
- Customers may run Maestro on Cadence’s cloud, any public cloud, or fully on-premises and may fine-tune the models on their own designs and workflows to keep data private.
- Cadence said Maestro is built in part on Nvidia’s Nemotron and uses Nvidia’s OpenShell safety tooling as part of a deepening partnership with Nvidia.
- The company reported early internal results showing up to 30% higher design quality and up to five times better token or cost efficiency versus general-purpose models and said select customers get early access in Q4 2026.
- Cadence executives say rising costs from repeated agent calls make it sensible to encode more iteration into model weights, a shift that could lower runtime costs and shift industry value toward vendors that control models and fine-tuning.