Ricursive Intelligence pitches AI that designs its own chips — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Ricursive Intelligence founders Anna Goldie and Azalia Mirhoseini pitched at TechCrunch Disrupt on Wednesday, making the case that AI-designed chips can address a coming compute crunch.
- Anna Goldie said compute demand will grow at least 10,000x in the next five years, called that estimate "potentially conservative," and argued efficiency must come from chips rather than new data centers.
- The startup's AI designs chips, learns from each iteration, and improves subsequent designs, aiming to automate and accelerate the chip design process end-to-end.
- AlphaChip, an earlier system from the same founders, produced a circular "donut-shaped" chip component arrangement that initially surprised engineers but ultimately proved successful.
- The founders said they optimize entire chips holistically rather than partitioning them into dozens of smaller blocks, a method they said produces intricate layouts including wave-like patterns they plan to detail later.
- Goldie argued that compressing chip design from one to two years to roughly a week would make custom chips financially viable for space, AR/VR, and biological computing applications.
Why it matters: Chip design today runs one to two years per part, which limits custom silicon to mass-market products; Goldie's pitch is that collapsing that cycle to a week would unlock custom chips for niche applications like space, AR/VR, and biological computing. Goldie framed an expected 10,000x compute demand increase as "potentially conservative" and pointed to chip efficiency, not new fabs or data centers, as the lever to pull.
Ask SkimNews
