Ricursive Intelligence’s AI-designed chip bet aims at the compute crisis

Ricursive Intelligence founders Anna Goldie and Azalia Mirhoseini described a system that can design chips, learn from each iteration, and improve future designs. Their goal is to automate and accelerate a process that traditionally requires careful human planning and long development cycles.

The practical takeaway: Ricursive is positioning chip design itself as an AI workflow, not just a hardware engineering task.

Goldie said onstage that demand for compute could rise at least 10,000x over the next five years, calling that estimate potentially conservative. The founders argue that the answer should not simply be building far more data centers, but designing chips that use data center bandwidth, wafers, and fabs more efficiently.

That frames AI-designed chips as a potential efficiency play in the broader compute crunch.

The founders said AI can challenge traditional chip design methods by finding layouts engineers may not expect. They pointed to AlphaChip, an AI system they previously developed, which produced a circular, “donut-shaped” arrangement of chip components that surprised engineers but ultimately worked.

Ricursive says it is optimizing entire chips rather than breaking them into dozens of smaller blocks, enabling intricate designs such as wave-like patterns.

Goldie argued that if new chips take six months, a year, or two years to design, teams have to be highly selective about what they build. If design cycles shrink dramatically, she said, custom chips could become financially practical for more applications, including space, AR, VR, and biological computing.

The broader implication is clear: faster chip design could expand the range of specialized hardware teams are willing to pursue.