- Published on
Ricursive Intelligence is betting on AI-designed chips to help solve the compute crisis
- Authors
- Name
- DeepHorus
Ricursive Intelligence founders Anna Goldie and Azalia Mirhoseini unveiled their startup’s plan at TechCrunch Disrupt on Wednesday. The company is building AI that designs computer chips, learns from each iteration, and improves later designs. The goal is to automate and accelerate chip design amid a worsening compute crisis.
Goldie said compute demand could grow by at least 10,000 times in five years. She argued that building 10,000 times more data centers is not sustainable. Ricursive instead aims to make chips more efficient, using bandwidth, wafers, and fabs more effectively.
The founders say AI can discover unconventional chip layouts that outperform traditional ones. Their earlier system, AlphaChip, produced a circular “donut-shaped” arrangement that initially surprised engineers. It ultimately proved successful, showing AI’s potential for nonintuitive solutions.
Ricursive is optimizing entire chips rather than partitioning them into dozens of smaller blocks. This lets AI create intricate patterns, including wave-like designs, which the founders plan to share later. Full-chip optimization could unlock new performance and energy efficiency.
Goldie said AI-designed chips will make many previously impractical chips possible. If designing a chip takes six months to two years, companies must be highly selective. But if design takes a week, custom chips for space, AR, VR, and biological computing become financially viable.
Some may fear AI designing its own chips, Goldie acknowledged. She argued that letting AI handle execution frees humans to focus on creative ideas and new applications. This shift is already happening in chip design, software development, and elsewhere.
Ricursive’s bet reflects a broader race to solve compute bottlenecks. By combining AI-driven design with faster iteration, the startup hopes to expand what computing can accomplish. The founders said they will reveal more about their wave-like patterns and full-chip approach in the future.
