Ricursive Intelligence is developing AI tools that design computer chips, evaluate each iteration and use what they learn to improve the next one. Founders Anna Goldie and Azalia Mirhoseini say the approach could help meet rising demand for computing power without relying solely on building more data centers.
Goldie expects computing demand to grow dramatically over the next five years. She argues that more efficient chips could help make better use of data-center bandwidth, manufacturing wafers and fabrication facilities.
From chip blocks to complete designs
Rather than dividing a chip into many smaller sections for separate optimization, Ricursive's approach aims to improve the design as a whole. The founders say AI can explore unusual layouts that conventional methods might miss. One example is AlphaChip, a system they previously developed that created a circular arrangement of chip components--an unconventional design that proved effective.
The founders believe faster, automated design could also make custom chips more practical. If development takes weeks instead of months or years, specialized processors may become more viable for fields including augmented and virtual reality, space applications and biological computing. The team also says these chips could be designed with energy efficiency in mind.
Goldie views AI as a way to handle more of the execution work, giving people greater room to focus on ideas, applications and new uses for computing. If the approach delivers on its promise, AI-assisted chip design could broaden access to specialized hardware and support more efficient computing in the years ahead.
