Ricursive Intelligence bets on AI-designed chips for compute crunch

Ricursive Intelligence founders Anna Goldie and Azalia Mirhoseini took the stage at TechCrunch Disrupt on Wednesday to detail how their startup is developing AI that can design chips, learn from each iteration, and improve subsequent designs, with the goal of automating and accelerating the chip-design process. The pitch rests on a demand forecast Goldie laid out on stage: “There is this understanding that in the next five years we are going to have at least 10,000x more demand for compute, and that’s somewhat potentially conservative,” she said. “So we don’t want to build 10,000 times more data centers. We think that we need to help by making chips that are more efficient, that use the bandwidth in those data centers more efficiently, that use the wafers and fabs more efficiently, because even if we make [data centers] in space or something, I don’t know, you won’t see the sun anymore.”

Some may be unsettled by the idea of AI designing its own chips, but Goldie argues that handing the task to AI frees humans to focus on building applications and exploring new possibilities for what computing can do. She said the same shift is already underway in chip design, software development, and elsewhere: humans supply the ideas, and AI increasingly handles the execution.

The founders say AI is challenging conventional chip design by discovering unconventional layouts that outperform traditional ones. AlphaChip, the AI system the pair previously developed, produced a circular, “donut-shaped” arrangement of chip components that initially surprised engineers but ultimately proved successful. At Ricursive, they are optimizing entire chips at once rather than following the traditional approach of partitioning a chip into dozens of smaller blocks — an approach they said lets the AI create intricate layouts, including wave-like designs, that they plan to detail further in the future.

The founders argue that AI-designed chips will make it possible to build kinds of chips that previously weren’t feasible to design at all. If designing a new chip takes six months, a year, or two years, Goldie said, a company has to be extremely selective about which chips it even attempts. Cutting that to roughly a week, she said, would make it financially viable to design custom chips for applications such as space, AR and VR, and biological computing — and to make those chips more energy efficient.

Why it matters

If AI-designed chips genuinely cut design cycles from years to roughly a week, as Goldie suggests, the barrier to building specialized silicon for niche workloads — from AR/VR headsets to biological computing — could fall dramatically, reshaping who can afford to design custom hardware rather than just license it.