Chips are getting more powerful to support AI, which means they need to be designed with ever greater precision. That design work, however, is slow and difficult. The question then becomes what happens when AI begins to take part in that design labor as well.
The concept behind Ricursive Intelligence involves AI that designs chips, learns from the experience, and then uses that knowledge to enhance future designs. The startup is building this capability in the hope of speeding up the development of hardware that powers increasingly capable AI.
Today, the process of designing a chip stretches across two to three years. Recursive’s aim is to compress that entire cycle into just a matter of weeks.
At TechCrunch Disrupt 2026, Ricursive Intelligence co-founders Anna Goldie and Azalia Mirhoseini will take the Disrupt Stage for “When AI Starts Designing Its Own Hardware.” They’ll discuss closing the loop between AI and chip development, building self-improving systems, and tackling one of the biggest bottlenecks in the race toward more powerful intelligence.
From AlphaChip to a $4 billion startup
The pair’s work with AlphaChip at Google has already shown the power of AI in chip design. Before founding Ricursive Intelligence, Goldie and Mirhoseini co-led the project, which generated chip layouts in hours — a far faster process than what human designers require. Their system helped design multiple generations of Google’s Tensor Processing Units.
Ricursive Intelligence is pushing the idea further, developing AI tools that automate and speed up chip design. The system is meant to learn across different chips, meaning the experience of designing one can make it better at handling the next.
The idea is that AI can help create better hardware, which in turn supports more powerful AI, and those systems can then help develop what comes next — a loop that could become very strong in practice.
The founders behind the project
Google’s ML for Systems team was co-founded by Goldie and Mirhoseini. The two were also early employees at Anthropic, and held the senior staff research scientist position at Google DeepMind.
Goldie is an Innovators Under Ricursive’s founder and CEO, holds a Ph.D. in computer science from Stanford and was named one of MIT Technology Review’s 35 Innovators Under 35, and Ricursive’s Mirhoseini is its founder and CTO. Mirhoseini is also an assistant professor of computer science at Stanford, where he founded its Scaling Intelligence Lab.
Ricursive was launched in late 2025, and money came in fast. The company raised $335 million within four months, closing a Series A at a valuation of $4 billion, including a $300 million round. Nvidia is one of its backers.
How the AI-hardware loop works
The hardware beneath AI models follows a distinct development timeline, and Ricursive is betting that AI itself can help shorten it. The company’s approach relies on a particular sequence of steps.
Goldie and Mirhoseini developed the system to handle more of the complex chip-design process, including component placement and design verification, while learning from one design to the next. They think this automation will speed up chip design and open the door to new chip architecture, which they believe could produce more capable and efficient AI.
What the session covers
Researchers Goldie and Mirhoseini contributed to demonstrating that AI can design real chips for the world, and they are now founding a company to push that idea even further. Founders, investors, and technology leaders have an opportunity to grasp why the pace of chip creation could come to define what AI can accomplish next.
The conference spans six industry stages, roundtables, and breakouts at Disrupt, October 13–15 at Moscone West in San Francisco, and their session is among 200+ of them. The event draws more than 10,000 founders, investors, operators, and tech leaders, along with 250+ speakers and 300+ exhibiting startups. The schedule is only part of the draw; matchmaking, dealmaking, and networking offer additional chances to connect with the founders, investors, and builders shaping what comes next.
Key dates and details
| Date | Event |
|---|---|
| Late 2025 | Ricursive founded by Goldie and Mirhoseini |
| Four months later | $335 million raised at $4 billion valuation, including $300 million Series A |
| October 13–15 | TechCrunch Disrupt 2026, Moscone West, San Francisco |
The stakes of faster chip design
The bottleneck in AI capability isn’t just software. It’s also the hardware that runs it. Faster chip design could accelerate the entire field.
Ricursive has set its sights on building a system that learns across chip designs and applies that knowledge to the next one, rather than treating chip design as a fixed, manual task.
The real issue is whether AI can keep up with its own requirements. Should Ricursive succeed, the loop between AI and hardware might become self-sustaining — each generation of AI improves the next generation of hardware, which then enables even more powerful AI.
Ricursive’s leaders will speak at Disrupt, giving the audience a direct look into the technology, bottlenecks, and self-improving systems that are central to their work. Goldie and Mirhoseini will handle the discussion.
The stakes are high. The bottleneck in AI capability isn’t just software. It’s also the hardware that runs it. Faster chip design could accelerate the entire field.
Anyone paying attention to the field of AI has reason to watch closely now. The apparatus behind AI is on the verge of gaining the ability to improve itself.
Source material: “TechCrunch Disrupt 2026: Ricursive Intelligence’s Anna Goldie and Azalia Mirhoseini on when AI starts designing its own hardware,” TechCrunch.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.