Ricursive Intelligence founders Anna Goldie and Azalia Mirhoseini announced at TechCrunch Disrupt on Wednesday that their startup is developing artificial intelligence systems to design more efficient computer chips and alleviate the global compute crisis. The company's technology aims to automate and accelerate the chip design process, allowing AI to learn from each iteration and improve subsequent hardware layouts.
Addressing the projected 10,000-fold increase in compute demand over the next five years, the founders emphasized that building more data centers is unsustainable. Instead, they are using AI to discover unconventional layouts that outperform traditional designs, such as circular and wave-like patterns. By optimizing entire chips rather than partitioning them into smaller blocks, the system can create intricate arrangements that traditional methods overlook.
This automated approach could reduce chip design timelines from years to weeks, making custom silicon financially viable for specialized applications like space exploration, augmented reality, and biological computing. The founders argue that shifting execution to AI frees humans to focus on creative ideas and application development. This hardware innovation matters because it addresses surging computational demands while drastically improving energy efficiency and manufacturing timelines for next-generation technology.