AI Chip Design Startup Eyes Global Optimum as Ricursive Challenges Hardware-Constrained Architecture

Ricursive Intelligence, the startup founded by former Google researchers Anna Goldie and Azalia Mirhoseini, is unveiling an AI system that designs silicon chips in weeks rather than years—raising a provocative question: what if current AI architectures were shaped not by theoretical superiority but by the hardware constraints of their era? At TechCrunch Disrupt 2026 in September, the co-founders presented their vision of closing the feedback loop between AI and chip design. Ricursive's thesis is that AI can be trained to autonomously generate chip layouts, learn from those designs, and improve subsequent iterations—establishing what they call a recursive self-improvement cycle.