Ricursive Intelligence Aims to Cut AI Chip Design From Years to Weeks
Ricursive Intelligence aims to cut chip design from years to weeks with self-improving AI, backed by $335 million and a $4 billion valuation from investors.
Summary
Ricursive Intelligence is building AI to compress chip development from two to three years to weeks, automating work from component placement through design verification and learning across chips so each project improves the next. Co-founders Anna Goldie and Azalia Mirhoseini previously co-led Google’s AlphaChip, which generated layouts in hours and helped design multiple generations of Google Tensor Processing Units. Their proposed feedback loop uses AI to create better hardware, which supports stronger AI that can accelerate subsequent designs and enable more capable, efficient systems.
Founded in late 2025, Ricursive raised $335 million within four months at a $4 billion valuation, including a $300 million Series A, with Nvidia among its investors. Goldie, founder and CEO, and Mirhoseini, founder and CTO, also co-founded Google’s ML for Systems team, were early Anthropic employees and became senior staff research scientists at Google DeepMind. Goldie holds a Stanford computer science Ph.D. and received a 35 Innovators Under 35 honor from MIT Technology Review. Mirhoseini is a Stanford computer science assistant professor and founded its Scaling Intelligence Lab.
Goldie and Mirhoseini will discuss self-improving systems, chip development bottlenecks and Ricursive’s approach at Disrupt, October 13 to 15, 2026, at Moscone West in San Francisco.
Positives
- Ricursive aims to reduce chip development from two to three years to weeks.
- AlphaChip generated chip layouts in hours and helped design multiple generations of Google Tensor Processing Units.
- Ricursive raised $335 million within four months, including a $300 million Series A, at a $4 billion valuation.
- Nvidia’s investment gives Ricursive backing from a leading AI chip company.
- Ricursive plans to automate work from component placement through design verification while learning across successive chips.
Risks & concerns
- Ricursive’s target of reducing chip design to weeks remains a goal rather than a demonstrated result.
- Current chip development takes two to three years, leaving hardware on a slower cycle than rapidly advancing AI models.
- AlphaChip proved the approach at Google, but performance details for Ricursive’s independent system remain limited.
- Automating placement through verification gives Ricursive a broad and technically complex development mandate.