Ricursive Intelligence uses AI to accelerate chip design

Ricursive Intelligence is tackling one of the biggest bottlenecks in artificial intelligence by using AI to design the very hardware that powers it. Founded in late 2025 by former Google DeepMind researchers Anna Goldie and Azalia Mirhoseini, the startup aims to compress the traditional chip design cycle from two to three years down to just a few weeks. The company's rapid rise caught the attention of major industry players, securing $335 million in funding at a $4 billion valuation within its first four months, which included a $300 million Series A round with participation from Nvidia.

The co-founders bring deep expertise to this challenge, having previously co-led the development of AlphaChip at Google. That system demonstrated that AI could generate complex chip layouts in hours rather than the months required by human designers, ultimately helping design multiple generations of Google's Tensor Processing Units. At Ricursive, Goldie and Mirhoseini are expanding this concept to automate the entire pipeline, from initial component placement through design verification. Their system is designed to learn continuously, using the experience of designing one chip to improve how it approaches the next.

For hardware engineers and AI practitioners, this technology could fundamentally alter the pace of hardware iteration. Instead of waiting years for specialized silicon to catch up with rapidly evolving software architectures, developers could soon deploy custom chips optimized for specific algorithmic breakthroughs in a fraction of the time. This creates a self-improving feedback loop where advanced AI designs superior hardware, which in turn runs more powerful AI models. Goldie and Mirhoseini will discuss these developments and the future of self-improving hardware systems at the upcoming TechCrunch Disrupt 2026 conference, taking place October 13-15 at Moscone West in San Francisco.