Ricursive Intelligence: A step-change on the path to AGI

When we talk about AI progress, we tend to talk about models. But models aren’t what’s holding the field back anymore. The real constraint is physical: the chips that run these systems—and the time and cost it takes to design and manufacture them—are becoming the limiting factor on how fast AI can scale.

Designing leading-edge silicon takes years and hundreds of millions of dollars. Traditional EDA tools struggle to keep up with the complexity of modern accelerators. And yet, demand has never been higher. The global semiconductor market is set to approach $1 trillion in 2026. Every frontier model, every hyperscaler, and every new AI-native product ultimately runs into the same constraint: how quickly—and how affordably—we can design the chips underneath it all. The future of the chip industry depends on new designs.

Ricursive Intelligence is a step-change on the path to AGI. And we truly believe there’s no one better poised to solve this problem than co-founders Dr. Anna Goldie and Dr. Azalia Mirhoseini. Previously, they worked at Google DeepMind and co-created AlphaChip, a novel reinforcement learning method for chip design, credited with transforming how Google designs TPUs. 

Ricursive extends that foundation into something bigger. At Ricursive, the loop between AI and hardware closes: models design chips, those chips accelerate better models, and the system continually improves itself. In a field obsessed with abstract definitions of AGI, Ricursive has actually presented a concrete path to get there. 

Today, we announce that Felicis has formalized our partnership with this extraordinary team by participating in Ricursive’s Series A. But we’ve been lucky to know these founders for some time. Michelle’s relationship with Azalia goes back years. Felicis is a SEAMS affiliate at Stanford, and by sponsoring her Scaling Intelligence Lab there, we’ve stayed close to the research questions that ultimately led to the founding of Ricursive.

Separately, James was lucky enough to score a seat next to Anna at a research symposium. He knew immediately that this was someone he wanted to work with and support. Anna’s obviously technically brilliant, but also curious, principled, and utterly-clearsighted about the future. 

Anna and Azalia began exploring ideas together in 2018. They kept coming back to the same thesis: chip design had become the primary bottleneck in AI scaling, and model-hardware co-evolution was the most credible path forward. Watching them work together, it’s clear why they gravitated towards each other. They have a deep history of collaboration, mutual respect, and an almost uncanny ability to think in sync. It’s the kind of co-founder dynamic you can’t manufacture. By the time Ricursive formally came together, our decision process was simple. The trust was already there.

Ricursive is a perfect example of our researcher-founder thesis: when you give world-class scientists the space for deep, unconstrained research instead of forcing early commercialization, you don’t just get a company. You get a new technical frontier. We believe the next wave of AI progress won’t come from incremental product tweaks on today’s language models. It’ll take researcher-founders like Azalia and Anna who are willing to do the hard, long-horizon work to get us there.