Ricursive Intelligence's founders, Anna Goldie and Azalia Mirhoseini, are betting that smarter, more efficient AI chip design can be a key part of solving the growing compute crisis. Goldie laid out the scale of the challenge during a stage appearance, noting that current projections suggest demand for compute could grow by at least 10,000x over the next five years—a figure she actually considers somewhat conservative. Rather than trying to build 10,000 times as many data centers, she argued, the better path is to make chips that squeeze more out of existing infrastructure, using bandwidth, wafers, and fabs far more efficiently. As she put it, even if data centers were somehow built in space, the result would be a world where "you won't see the sun anymore."

Some people may find the idea of AI designing its own chips unsettling, but Goldie sees it differently. In her view, handing chip design over to AI frees humans to concentrate on building applications and chasing new ideas that push the boundaries of what computing can do.

"I think the way everything is shifting is that humans come bring their great ideas and creative thoughts, and execution is handled by AI," Goldie said. "And this is happening in chip design, software development, and other places. And that's something that recursive does for chip design."

According to the founders, AI is upending conventional chip design by uncovering unconventional layouts that outperform traditional ones. One striking example came from AlphaChip, an AI system the founders built earlier, which generated a circular, "donut-shaped" arrangement of chip components. Engineers were initially taken aback, but the design ultimately proved successful. The pair explained that instead of the traditional method of splitting chips into dozens of smaller blocks, they optimize entire chips as a whole. That lets AI produce complex patterns, including wave-like designs, which they say they'll detail further in the future.

The founders also contend that AI-designed chips will unlock entirely new categories of chips that simply weren't feasible before. "If it takes six months or a year or two years to design a new chip, then you really have to be extremely selective about the chips that you design," Goldie said. "But if you can design a chip in like a week, and then say for manufacturing, you can design custom chips for all sorts of different types of applications in like space and AR, VR, and you biological computing, and it makes sense financially to do that, and we can be more energy efficient."