# Ricursive Develops AI Software to Design and Refine Chip Layouts By Simon Yoon Canonical URL: https://www.tokenpost.com/news/technology/32164 Published: 2026-10-15T01:27:38.000Z Updated: 2026-10-15T01:27:38.000Z Section: Technology > Co-founders Anna Goldie and Azalia Mirhoseini say more efficient chips could help meet Goldie’s projection that computing demand will rise at least 10,000-fold in five years. Ricursive Intelligence is developing AI software to design and refine chip layouts. Goldie said a process that takes about a week could make custom chips practical for more uses. Co-founders Anna Goldie and Azalia Mirhoseini discussed the effort at an event in San Francisco Wednesday. Goldie projected that demand for computing could grow at least 10,000-fold over the next five years, describing that estimate as potentially conservative. “We don’t want to build 10,000 times more data centers,” she said. The founders argue that more efficient chips and better use of data-center bandwidth could help meet that demand. They also point to using wafers and fabrication facilities more efficiently. Goldie framed AI automation as a way to handle more of the execution work, leaving people room to develop ideas and applications. Ricursive’s software is being built to improve chip designs through repeated iterations. The founders’ approach evaluates a whole chip rather than breaking its layout into dozens of smaller blocks. They say that broader view can help the system find unconventional arrangements and intricate patterns. One example they cited is AlphaChip, an earlier system they developed. It produced a circular arrangement of chip components that surprised engineers before proving successful. The founders plan to share more about layouts with wave-like patterns. Goldie said chip projects that take six months, a year or as long as two years force companies to be highly selective about what they design. She said a process that takes about a week could make custom chips financially practical for more uses, including space, augmented and virtual reality, and biological computing. She also said such designs could improve energy efficiency. A U.S. 1-gigawatt AI data center can require about [$38 billion in upfront capital when chips are included](). ## Links in this article - [$38 billion in upfront capital when chips are included](https://www.tokenpost.com/news/technology/30781)