Tuesday, 17 November 2026

AI-designed chips could cut data center costs and delays

Ricursive Intelligence, a $4 billion startup from the creators of Google's AlphaChip, wants AI to design chips in a week instead of a year or more.

Tech Crunch Disrupt, related to: AI-designed chips could cut data center costs and delays
Photo: TechCrunch / CC BY 2.0 (modified)

The short version

  • Ricursive Intelligence is building AI that designs computer chips and learns from each attempt to make the next design better.
  • Its founders created AlphaChip at Google, an AI tool that produced chip layouts in hours and helped design Google's TPU chips.
  • The startup raised a $300 million Series A at a $4 billion valuation and counts Nvidia's venture arm among its investors.
  • In November it released Ripple-65B Preview, an open coding model it says is a variant of the AI behind its chip platform.
Quick read · 2 min

Ricursive Intelligence wants AI to design computer chips, shrinking a process that can take six months to two years down to about a week. Its founders, Anna Goldie and Azalia Mirhoseini, built Google's AlphaChip, which produced chip layouts in hours and helped design Google's TPU chips.

The startup doesn't make chips. It builds design tools for chipmakers like Nvidia, AMD and Intel. Faster design could make custom, energy-efficient chips affordable for narrow jobs, easing the strain AI puts on data centers.

  • It raised a $300 million Series A at a $4 billion valuation.
  • Nvidia's venture arm is an investor.
  • In November it released Ripple-65B Preview, an open coding model.

No chip designed end to end by its platform has been announced yet, so the big claims still need real silicon to back them up.

Designing a new computer chip can take a team of engineers a year or more. Anna Goldie and Azalia Mirhoseini, the founders of Ricursive Intelligence, say AI-designed chips could bring that down to about a week, and they think the world will need that speed sooner than most people expect.

Why should you care about how chips get drawn up? Because chips are now the bottleneck behind almost everything AI does, from the chatbot on your phone to the giant data centers that run it. Goldie said on stage at the Disrupt tech conference in October that demand for computing could grow at least 10,000 times over the next five years, and called that estimate possibly conservative. Building that many more data centers isn’t realistic, she argued. Making chips that do far more with the same power and the same factories is the alternative.

01

What Ricursive Intelligence actually does

Ricursive builds AI tools that design chips. It does not make the chips itself. That sets it apart from most AI chip startups, which want to compete with Nvidia. Here, Nvidia is an investor, and chipmakers like Nvidia, AMD and Intel are the customers Ricursive is aiming at.

The system works in a loop. The AI produces a chip design, learns from how that attempt performs, and uses those lessons to improve the next one. The company describes its goal as closing the feedback loop between AI models and the chips that run them, so better AI produces better chips, which in turn power better AI.

Mirhoseini has said the aim is for any kind of chip, from custom designs to more traditional parts, to be built in an automated and much faster way.

Close-up of advanced chip testing equipment in a lab
02

Why the founders’ AlphaChip work matters

This is not a first attempt. At Google, the pair created AlphaChip, an AI tool that could produce workable chip layouts in hours, a job that normally takes human designers a year or more. AlphaChip was used to help design Google’s Tensor Processing Units, the custom AI chips Google runs in its data centers. Accounts differ on how many generations: the company’s funding announcement says four, while an earlier profile said three.

One AlphaChip result became famous inside the field. It arranged chip components in a circular, donut-shaped layout that surprised engineers at first but went on to work well. The founders say this is the point: AI can find layouts humans would never try.

Ricursive is now going further. Instead of the usual method of splitting a chip into dozens of smaller blocks and optimizing each one, it works on the whole chip at once. The founders say that lets the AI produce intricate patterns, including wave-like designs they plan to share more about later.

The work has not been free of controversy. One Google colleague spent years trying to discredit the AlphaChip research and was fired in 2022, according to an earlier report.

03

How AI-designed chips could cut data center costs

The biggest promise is time, not tiny gains in speed. Goldie’s argument is that when a chip takes six months to two years to design, companies have to be extremely picky about which chips they make at all. If design takes about a week, it becomes financially sensible to make custom chips for narrow jobs.

She listed examples such as chips for space, for augmented and virtual reality, and for biological computing. Custom chips tuned to one job usually waste less energy than general ones, which is where the data center savings would come from: more work from each chip, each wafer and each factory.

For ordinary people, the payoff would be indirect but real. Cheaper, more efficient chips could mean AI services that cost less to run, data centers that draw less power from local grids, and gadgets with chips built for exactly what they do. None of that is proven yet. Ricursive has not announced a chip designed end to end by its platform reaching customers.

04

Big money and a surprise coding model

Investors are betting heavily anyway. Sequoia Capital led a $35 million seed investment when Ricursive launched in December 2025. Less than two months later, in January 2026, Lightspeed Venture Partners headed a $300 million Series A that valued the company at $4 billion. Other backers include DST Global, Felicis and NVentures, Nvidia’s venture capital arm. The founders previously worked together at Google Brain and were early employees at Anthropic.

Then, on November 4, Ricursive released something unexpected: Ripple-65B Preview, an open-weight AI model for writing code. Anyone can download it under the permissive Apache 2.0 license. The company says it is a variant of the private model that powers its chip-design platform, which makes sense because chip design is heavily software work.

Ricursive claims Ripple-65B scored 80.5 on the Terminal-Bench 2.1 coding test, just ahead of a much larger rival from Reflection AI, and says the model cost about $32 million in total research and development. Those are the company’s own results, and its charts show bigger models still ahead in several other areas.

05

Quick answers

What are AI-designed chips?

They are chips whose layout is planned by AI software instead of mainly by human engineers. The AI tries designs, learns what works and improves with each round.

Does Ricursive Intelligence make its own chips?

No. It builds AI design tools and wants chipmakers such as Nvidia, AMD and Intel to use them.

Who founded Ricursive Intelligence?

Anna Goldie, the CEO, and Azalia Mirhoseini, the CTO. They created Google’s AlphaChip and worked at Google Brain and Anthropic.

You're reading the quick version.