Ricursive Intelligence is using AI to rethink chip design

Ricursive Intelligence is developing AI that can design chips, learn from each iteration and improve later designs. The startup’s founders, Anna Goldie and Azalia Mirhoseini, described the approach at TechCrunch Disrupt, saying their goal is to automate and speed up chip design.

Goldie said demand for computing power could rise by at least 10,000x in the next five years, a projection she described as potentially conservative. She argued that more efficient chips could help meet that demand without building 10,000 times more data centers. In her account, the opportunity includes using data-center bandwidth, wafers and fabrication facilities more efficiently.

The founders said AI can find unconventional chip layouts that may outperform traditional designs. They pointed to AlphaChip, an AI system they previously developed, which produced a circular, “donut-shaped” arrangement of chip components that surprised engineers and ultimately proved successful. Ricursive says it is optimizing entire chips instead of dividing them into dozens of smaller blocks. The founders said this can enable intricate patterns, including wave-like designs, and that they plan to share more about those designs in the future.

Goldie said designing chips in a week could make custom chips practical for a wider range of applications, including space, AR, VR and biological computing. She said faster design could also make it financially viable to develop chips for more applications and support greater energy efficiency. The founders presented these possibilities as goals for AI-assisted chip design.