How Two Former Google Brain Researchers Built a $4 Billion Chip Design

When Anna Goldie and Azalia Mirhoseini started Ricursive Intelligence four months ago, they weren’t just launching another AI chip startup. They were betting that the future of artificial intelligence depends on chips that can design themselves. Last month, that bet paid off spectacularly: a $300 million Series A round at a $4 billion valuation, led by Lightspeed, just weeks after a $35 million seed from Sequoia. The numbers are eye-popping, but the story behind them is even more remarkable.

“We want to enable any chip to be built in an automated and very accelerated way. We’re using AI to do that.” — Azalia Mirhoseini, CTO of Ricursive Intelligence

The A&A Origin Story

Goldie and Mirhoseini’s partnership began at Stanford, where Goldie earned her PhD while Mirhoseini taught computer science classes. Their careers have moved in perfect sync ever since. “We started at Google Brain on the same day. We left Google Brain on the same day. We joined Anthropic on the same day. We left Anthropic on the same day,” Goldie recounted. “Then we started this company together on the same day.”

At Google, the pair—nicknamed “A&A” internally—created something extraordinary: Alpha Chip, an AI tool that could generate high-quality chip layouts in six hours. Human designers typically spend a year or more on the same task. The tool went on to help design three generations of Google’s Tensor Processing Units, the specialized chips that power much of Google’s AI infrastructure.

Their reputation in the AI world is such that both founders received what Goldie described as “weird emails from Zuckerberg making crazy offers.” They declined. That pedigree explains why VCs were practically lining up outside their door before Ricursive even had a product.

Not Your Typical Chip Startup

The fundamental difference between Ricursive and nearly every other AI chip startup is that they’re not trying to build chips—they’re building the AI that designs them. This makes them partners rather than competitors to chip giants. Nvidia, AMD, and Intel aren’t threats; they’re potential customers. In fact, Nvidia is already an investor.

The technical challenge is staggering. Modern computer chips contain millions to billions of logic gate components packed onto silicon wafers. Placing those components to optimize performance, power efficiency, and thermal characteristics requires solving an incredibly complex spatial puzzle. Human engineers can spend months or years iterating on designs.

The Alpha Chip approach uses what Goldie calls “a reward signal” that rates design quality. The AI agent takes that rating and “updates the parameters of its deep neural network to get better.” After thousands of designs, the agent doesn’t just improve—it accelerates. Each chip it designs makes it a better designer for the next one.

“Chips are the fuel for AI. I think by building more powerful chips, that’s the best way to advance that frontier.” — Anna Goldie, CEO of Ricursive Intelligence

The Path to Self-Improving AI

Ricursive’s platform expands on the Alpha Chip concept in crucial ways. It incorporates large language models and handles the entire design pipeline—from component placement through design verification. Any company that makes electronics and needs custom chips is a potential customer.

The founders believe their technology could play a central role in achieving artificial general intelligence. The ultimate vision is almost poetic: AI designing the chips that power more advanced AI, in a continuously accelerating cycle. “We think we can enable this fast co-evolution of the models and the chips that basically power them,” Mirhoseini said.

For those wary of AI designing its own brains at ever-increasing speeds, the founders point to a more immediate benefit: hardware efficiency. When AI labs can design far more efficient chips, their growth won’t consume as much of the world’s energy and resources. Goldie estimates they could achieve “almost a 10x improvement in performance per total cost of ownership.”

What Comes Next

The startup isn’t naming its early customers yet, but the founders say they’ve heard from every major chipmaker. Given their track record and the caliber of their backers, they have their pick of development partners. The question now is whether they can execute at scale.

The chip design industry is notoriously conservative. Companies like Intel and TSMC have spent decades refining their processes. Convincing them to trust AI with billion-dollar designs won’t happen overnight. But if Ricursive delivers on its promises, the implications extend far beyond faster chip design—they could reshape the fundamental economics of computing.

For now, Goldie and Mirhoseini are focused on building. Four months in, they’ve already achieved what most startups never do. The real test starts now.

This article was reported by the ArtificialDaily editorial team. For more information, visit TechCrunch.