As artificial intelligence models grow more capable, the demand for specialized computing hardware is accelerating. Yet designing advanced chips remains a highly intricate process that can typically take two to three years.
Ricursive Intelligence, founded by AI researchers Anna Goldie and Azalia Mirhoseini, is developing systems designed to shorten that timeline dramatically. The company's goal is to use AI to automate key stages of chip creation, from component placement to design verification, with an ambition to reduce development cycles from years to weeks.
A learning loop for hardware innovation
The approach is built around a self-improving workflow: an AI system contributes to a chip design, learns from the outcome, and applies that experience to future designs. In theory, each completed project could make the next one more efficient.
Goldie and Mirhoseini previously helped lead AlphaChip, an AI-driven system used to generate chip layouts in hours. Their earlier work supported several generations of Google's Tensor Processing Units, demonstrating how machine learning can assist with real-world semiconductor engineering.
Ricursive Intelligence is expanding this concept beyond individual layouts. Its tools are intended to build transferable knowledge across multiple chip projects, enabling a broader design intelligence that evolves over time.
Bridging AI and semiconductor development
The founders will explore this emerging feedback loop at Disrupt 2026: better AI can help create more capable hardware, while more efficient hardware can power the next generation of AI systems. The discussion will focus on how this cycle may reshape the pace of technological development.
Launched in late 2025, Ricursive Intelligence quickly attracted substantial backing, including investment from Nvidia. Goldie, the company's CEO, holds a computer science doctorate from Stanford, while CTO Mirhoseini is a Stanford computer science professor and leads the university's Scaling Intelligence Lab.
If AI-assisted chip design becomes widely deployable, it could make advanced computing infrastructure faster to develop, more energy-efficient, and more adaptable to scientific and creative applications in the years ahead.
