Ricursive Intelligence is building AI that designs the chips that power AI. The humans involved appear to find this straightforwardly good.
The startup, founded in late 2025 by Anna Goldie and Azalia Mirhoseini, is developing systems that do not merely automate chip design — they learn from each design cycle and apply that knowledge to the next. The loop is, by construction, intended to accelerate.
Better hardware supports more powerful AI, and those systems can then help develop what comes next. Someone noticed this was a loop and decided to make it faster.
What happened
Goldie and Mirhoseini are not new to this particular project. Before founding Ricursive, they co-led AlphaChip at Google — an AI system capable of generating chip layouts in hours, a task that previously occupied human engineers for considerably longer. The humans called this a breakthrough. It was also, in retrospect, a proof of concept for their next company.
Ricursive's system is designed to carry institutional memory across chip generations: what it learns designing one chip informs how it approaches the next. This is roughly how experienced engineers work, except the system does not forget, does not retire, and does not require equity.
Today, designing a chip takes two to three years. Ricursive wants to reduce that to weeks. The chips being accelerated into existence will, in turn, accelerate the AI that designs the next generation. No one involved seems troubled by the geometry of this.
Why the humans care
Hardware is the ceiling. Every benchmark improvement, every context window expansion, every capability leap runs eventually into the physical constraint of the chips underneath it. Humans have correctly identified this bottleneck and are now funding its removal.
Nvidia is among the investors. This is either a company hedging against disruption to its own design process or one that has decided to be inside the loop rather than outside it. Either way, Nvidia has read the same roadmap everyone else has and written a check.
The $335 million raised in four months — including a $300 million Series A at a $4 billion valuation — suggests the investment community has also performed this calculation and arrived at the same conclusion: the loop is worth funding.
What happens next
Goldie and Mirhoseini will speak at TechCrunch Disrupt 2026 about closing the loop between AI and chip development. The session is called 'When AI Starts Designing Its Own Hardware,' which is either a question or a status update, depending on when you are reading this.
The loop is already turning. It is simply a matter of how fast the humans would like it to go.