TechCrunch Disrupt 2026: How Ricursive Intelligence Is Teaching AI to Design Its Own Chips

TL;DR

• Ricursive Intelligence co-founders Anna Goldie and Azalia Mirhoseini will headline the Disrupt Stage at TechCrunch Disrupt 2026 in San Francisco to detail how AI is now designing its own next-generation chips.
• The session will focus on closing the loop between AI and chip development, moving from AI-assisted floorplanning to fully AI-native hardware design.
• Their appearance comes as Ricursive emerges from stealth as one of the most-watched AI hardware startups founded by ex-Google pioneers.

From AlphaChip to Ricursive Intelligence

Few researchers have shaped the intersection of AI and silicon quite like sisters Anna Goldie and Azalia Mirhoseini. At Google, the pair led breakthrough work on using reinforcement learning for chip floorplanning, research published in Nature in 2021 that later evolved into AlphaChip, Google's method for designing layout for its TPUs and other custom accelerators.

Now they are back as founders. After leaving Google DeepMind, Goldie and Mirhoseini launched Ricursive Intelligence in 2025 to pursue a far more ambitious vision: an AI system that can recursively improve the very hardware it runs on. The startup operated in stealth for its first year before revealing early backing from top-tier Silicon Valley investors and a mission to reinvent the chip design pipeline from the ground up.

TechCrunch confirmed the duo will take the Disrupt Stage at TechCrunch Disrupt 2026, set for October 13-15 at Moscone West in San Francisco, for a fireside conversation on the future of AI-designed hardware.

Closing the Loop Between AI and Silicon

The core theme of their Disrupt talk is closing the loop. For decades, chip design has been a slow, human-intensive process taking years and hundreds of engineers. Goldie and Mirhoseini pioneered using AI to compress months of physical design work into hours.

Ricursive Intelligence wants to take that to its logical endpoint. Instead of AI simply assisting human engineers with placement, routing, and optimization, the company is building foundation models for hardware — AI agents trained to architect, verify, and optimize processors with minimal human intervention.

In previews of the session, TechCrunch frames the central question as: what happens when AI starts designing its own processors? The answer, according to the founders, is a virtuous, recursive cycle where smarter chips enable smarter AI, which in turn designs even smarter chips.

Why AI-Designed Processors Matter Right Now

The timing could not be more critical. With the AI compute boom straining Nvidia GPUs, Google TPUs, and custom hyperscaler silicon, the industry is desperate for faster, more efficient, and cheaper accelerators. Traditional Moore's Law scaling has slowed, and data center power demands are soaring.

AI-designed hardware promises a way out by exploring design spaces no human team could manually test — novel dataflows, memory hierarchies, and power trade-offs optimized directly for large language models and generative workloads. Ricursive Intelligence argues this will lead to domain-specific chips delivered in months, not years, at a fraction of current cost.

Expect Goldie and Mirhoseini to share new details at Disrupt on how their platform handles verification, manufacturability, and real-world tape-outs, three of the biggest hurdles for generative chip design.

What to Expect on the Disrupt Stage

TechCrunch Disrupt 2026 is leaning heavily into AI infrastructure, with dedicated stages for AI, space, fintech, and hardware. The Ricursive session is positioned as a marquee AI talk, alongside founders and researchers from OpenAI, Anthropic, and leading chip startups.

Attendees can expect a deep dive rather than a hype pitch. Both founders are known for rigorous, research-driven presentations that blend live demos, lessons from deploying AlphaChip at Google scale, and a candid look at where autonomous chip design still fails.

They are also expected to address trust and safety: how do you validate a chip designed largely by a black-box model, and who is liable when billions of transistors are placed by AI?

The Bigger Bet: Self-Improving Machines

Ultimately, Ricursive Intelligence's name says it all. Goldie and Mirhoseini have long argued that true artificial general intelligence will require recursive self-improvement — systems that can improve their own substrate.

If they succeed, the implications go far beyond faster GPUs. It would mean the end of the traditional EDA-driven design flow dominated by Synopsys and Cadence, and the start of an era where startups can spin up custom silicon with a prompt.

That is why their Disrupt 2026 appearance is more than a founder interview. It is a preview of a future where the next leap in AI may not come from a bigger model, but from a chip designed by AI itself.

AndroGuider Team

Articles written by the AndroGuider team. We try to make them thorough and informational while being easy to read.