Forget the celebrity guest list, TechCrunch Disrupt 2026 is about to get a lot more interesting. According to TechCrunch, Ricursive Intelligence co-founders Anna Goldie and Azalia Mirhoseini are taking the stage to discuss AI systems that design the chips that power them, then learn and improve from the process.
Chip design today takes two to three years. Ricursive wants to cut that to weeks. That's a fundamental change in how the hardware race operates.
Goldie and Mirhoseini are not newcomers to this space. Before launching Ricursive Intelligence as a $4 billion startup, they co-led AlphaChip at Google, an AI system that could generate chip layouts in hours instead of the months humans required. Their work shaped multiple generations of Google's Tensor Processing Units, the chips that power everything from search to AI models.
The core mechanism is a feedback loop: better AI designs better chips. Better chips support more powerful AI. That AI then designs even better chips. As AI models grow more ambitious, they need exponentially more powerful hardware. But the people designing that hardware cannot keep up with demand. Automating the designer means solving the chip design bottleneck, and potentially unlocking the next frontier of AI capability.
Goldie holds a PhD from Stanford, was named an MIT Technology Review 35 Under 35 innovator, and co-founded Google's ML for Systems team. Mirhoseini is Ricursive's founder and CTO, having previously worked as a senior staff research scientist at Google DeepMind and been an early employee at Anthropic.
The timing is sharp. The tech world is focused on the chip shortage narrative, and this conversation directly addresses why that shortage exists. For anyone interested in where AI hardware is heading, this matters.
