Computing

Ricursive Intelligence Bets on AI-Designed Chips

Founders Anna Goldie and Azalia Mirhoseini outlined their startup's mission to tackle the compute crisis by automating and accelerating chip design using artificial intelligence.

Radwan AbdulhadiEditor and technology writer

Published: October 15, 2026 · Updated: October 15, 2026 · 3 min read

Ricursive Intelligence Bets on AI-Designed Chips — Image: TechCrunch

Tackling the Growing Compute Crisis

As the technology sector looks toward future infrastructure challenges, startup founders Anna Goldie and Azalia Mirhoseini believe that more efficient chip design could help address the global compute crisis. Speaking at TechCrunch Disrupt, the creators behind [Ricursive Intelligence](https://techcrunch.com/2026/02/16/how-ricursive-intelligence-raised-335m-at-a-4b-valuation-in-4-months/) explained how their company is building artificial intelligence systems capable of designing chips, learning from every iteration, and continuously improving subsequent layouts to automate the entire process.

Goldie noted that industry expectations point to at least 10,000 times more demand for compute within the next five years, calling that projection potentially conservative. Rather than building thousands of times more data centers—which would introduce severe spatial and environmental hurdles—the startup wants to focus on making chips and data center bandwidth significantly more efficient.

Shifting Execution to Artificial Intelligence

While some industry observers might feel uneasy about artificial intelligence designing its own hardware, Goldie argues that this shift allows human engineers to concentrate on developing software applications and exploring new computational possibilities. According to the founders, human creativity brings the core ideas while execution is handled by intelligent automation.

This automated division of labor is already reshaping development workflows across multiple fields. The team emphasized that [AI](https://techcrunch.com/category/artificial-intelligence/) is successfully challenging traditional methodologies by uncovering unconventional component layouts that outperform standard configurations.

Unconventional Layouts and AlphaChip

Highlighting previous work, the founders pointed to AlphaChip, an intelligent system they developed earlier that famously produced a circular, donut-shaped arrangement of chip components. Although this layout initially surprised human engineers, it ultimately proved highly successful in practice.

Instead of following traditional partitioning methods that break a chip down into dozens of smaller blocks, the startup focuses on optimizing entire chips as unified systems. This strategy enables intelligent models to generate intricate patterns, including wave-like designs, which the founders plan to discuss in greater detail in the future.

Accelerating Custom Chip Development

The startup's approach could also democratize the creation of specialized hardware. Because designing a new chip traditionally takes anywhere from six months to two years, companies must be extremely selective about which architectures they pursue.

Goldie suggested that cutting the chip design cycle down to roughly a week would make it financially viable to build custom silicon for a wide range of emerging applications, spanning space technology, AR, VR, and biological computing, while maintaining higher energy efficiency.