Ricursive Wants AI to Design the Next AI Chips

Ricursive Intelligence is developing AI systems intended to automate large parts of chip design, creating a potential feedback loop in which AI helps build the hardware used to run future AI models.

What happened

The startup was founded by Anna Goldie and Azalia Mirhoseini, researchers associated with AlphaChip, Google's work on AI-assisted chip floorplanning. Ricursive is building systems that can generate chip designs, evaluate the results and use that feedback to improve subsequent designs. The company argues that automated design could eventually shorten development cycles and make specialised processors economical for more applications. Those timeline and productivity claims remain forward-looking, and large-scale commercial performance has not yet been independently demonstrated.

Why it matters

Chip design is complex, expensive and labour-intensive. Better automation could reduce engineering cycles and help hardware teams explore many more design alternatives than humans can evaluate manually. For AI companies, more efficient specialised chips could lower power use and inference costs.

The bigger picture

The strategic idea is recursive: better AI can improve chip design, better chips can provide more compute and that compute can support stronger AI. Whether that loop becomes commercially significant will depend on fabrication constraints, verification requirements and how much of the design process can safely be automated.