AI That Designs Its Own Chips: Ricursive's Anna Goldie and Azalia Mirhoseini

Neural networks are revolutionizing semiconductor design by achieving superhuman performance in layout optimization, a process previously reliant on thousands of human experts. Anna Goldie and Azalia Mirhoseini, co-creators of Google’s AlphaChip, are leveraging this technology through their new venture, Recursive Intelligence, to close the self-improving loop between AI and its physical substrate. The company aims to accelerate chip design by making tools like Static Timing Analysis engines 100,000x faster, eventually democratizing the field through a "design-less" model where custom hardware is tailored to specific workloads like frontier LLMs. By automating physical design and verification—stages that traditionally take years—they intend to reduce the massive opportunity costs associated with manufacturing delays. This approach enables a Cambrian explosion of specialized, high-performance chips that are more efficient and environmentally friendly than current general-purpose architectures.