Ricursive Intelligence, a startup founded by Anna Goldie and Azalia Mirhoseini, is building AI systems that design chips and improve on each iteration. At TechCrunch Disrupt, the founders argued that this is one way to address the looming compute crunch: rather than building 10,000 times more data centers to meet projected demand, they want to make chips, data-center bandwidth, and wafer/fab usage more efficient. Goldie called the 10,000x demand estimate over the next five years “somewhat potentially conservative.”

Their approach flips the traditional chip-design workflow. Instead of partitioning a chip into dozens of smaller blocks, the AI optimizes entire chips and discovers layouts a human engineer wouldn’t normally choose. As an example, they cited AlphaChip, their earlier system, which produced a circular, “donut-shaped” arrangement of chip components that surprised engineers but proved successful. The team says the same approach can generate wave-like patterns, which they plan to discuss later.

Goldie sees a broader shift in which humans contribute ideas and creativity while AI handles execution, not just in chip design but also in software development. Faster AI-driven design could change what gets built: if a chip can be designed in about a week instead of six months to two years, custom chips for space, AR/VR, and biological computing become financially viable and more energy-efficient. The article doesn’t disclose Ricursive’s current progress, funding, or technical pipeline.