Ricursive Wants to Entrust Chip Design to AI

Key Takeaways

1AlphaChip, developed by the founders of Ricursive, has generated a validated circular layout for a chip

2Ricursive is now optimizing entire chips, enabling AI to create complex patterns

3Anna Goldie anticipates a tenfold increase in computing demand over five years and advocates for shorter design cycles

Why it matters — AI chip design could accelerate hardware innovation and meet the rapidly growing computing needs.

Ricursive Intelligence highlights results and a method that shake up established practices: a previous system, AlphaChip, has already generated a validated circular layout, and the current approach aims at optimizing entire chips. The founders advocate that drastically shortened design cycles would pave the way for custom and more efficient chips. They position this strategy against a demand for computing that their co-founder Anna Goldie claims is set to increase by 10,000 times in five years.

A validated precedent and a holistic method

The founders cite a previous milestone: AlphaChip, developed earlier by the duo, produced a donut-shaped circular layout for chip components. This arrangement initially surprised engineers before proving effective. Ricursive now indicates it is optimizing entire chips rather than breaking them down into dozens of blocks, which would allow AI to generate more complex patterns, including wavy shapes. According to Anna Goldie and Azalia Mirhoseini, this ability to explore unconventional layouts can surpass traditional designs. They plan to elaborate on these patterns in the future.

The promise of short cycles and custom chips

Anna Goldie emphasizes that with cycles of six months, one year, or two years, the selection of chip projects remains drastic. She argues that reducing design time to one week would open the possibility of creating custom chips for various uses, from space to augmented and virtual reality, to biological computing. She believes such a scenario becomes financially viable and would also allow for energy efficiency gains. The founders more broadly assert that AI-designed chips would make previously inaccessible architectures feasible.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisis ton rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

The project: an AI that learns to design circuits

Founded by Anna Goldie and Azalia Mirhoseini, Ricursive Intelligence is developing an AI dedicated to chip design. The system is described as learning with each iteration to improve subsequent versions, with the goal of automating and accelerating the design phase. According to Anna Goldie, this philosophy of human input on ideas and execution entrusted to AI is precisely what the company applies to hardware design.

Justification: the demand for computing and the use of existing resources

Anna Goldie anticipates, over the next five years, at least 10,000 times more demand for computing, a projection she presents as possibly conservative. She does not aim for a massive deployment of data centers on the same scale but advocates for the design of more efficient chips that better utilize bandwidth, wafers, and fabs. She also argues that delegating design to AI would free humans for application development and the exploration of new possibilities. According to her, humans bring ideas and creativity, while AI ensures execution, a distribution she sees at work in chip design, software development, and beyond. These positions were presented on stage during TechCrunch Disrupt, on a Wednesday.