Ricursive Intelligence Raises $300M to Build AI That Designs Chips

Ricursive Intelligence just raised $300 million at a $4 billion valuation—55 days after announcing a $35 million seed round at $750 million. That’s a 5.3x valuation jump in under two months. The startup, founded by the creators of Google DeepMind’s AlphaChip, is building AI systems that design semiconductor chips. The thesis is elegantly recursive: AI designs better chips, better chips power better AI, and the loop accelerates.

“The pace of AI progress is dictated by hardware,” CEO Anna Goldie stated in the announcement. “Ricursive’s mission is to radically accelerate chip design, and ultimately to use AI to design its own silicon substrate.”

The Series A was led by Lightspeed Venture Partners, with participation from DST Global, NVentures (NVIDIA’s venture arm), Felicis Ventures, and Sequoia Capital. When NVIDIA invests in an AI chip design company, it’s worth paying attention.

AlphaChip: The proof that this works

Goldie and co-founder Azalia Mirhoseini aren’t pitching a theory. They built AlphaChip at Google DeepMind, and it’s been used in production for four generations of Google’s TPU chips—including the v5e, v5p, and Trillium. The technology also powers Google’s Axion Processors and MediaTek’s Dimensity 5G chips.

The results speak for themselves. AlphaChip generates chip floorplans in under six hours. Human experts typically take weeks to months for the same task. The AI-generated layouts match or exceed human designs in power efficiency, performance, and chip area. Google published the research in Nature in 2021, and open-sourced the code in 2022.

The technical approach treats chip floorplanning as a game—similar to how AlphaGo learned to play Go. A reinforcement learning system places components on the chip, receives feedback on the design quality, and iterates. The key innovation is an “edge-based” graph neural network that learns relationships between chip components, enabling the AI to understand how placement decisions affect overall performance.

The chip design bottleneck

Here’s why this matters: chip design is one of the most severe bottlenecks in AI progress. At leading-edge nodes (5nm, 3nm), designing a single chip takes 18-36 months and costs $450-650 million. Human labor accounts for 50-70% of that cost. Meanwhile, AI capabilities advance faster than hardware can keep up—creating an asymmetric gap that limits the entire field.

The EDA (Electronic Design Automation) market is dominated by two companies: Synopsys and Cadence, which together control over 60% of the roughly $20 billion market. Both are building AI features into their tools, but neither has production-proven AI chip design at the scale of AlphaChip. Sequoia’s investment thesis explicitly frames Ricursive as a potential disruptor to this oligopoly.

“Today, chip design takes 2-3 years and requires thousands of human experts,” Goldie said. “We will reduce that to weeks.”

The recursive self-improvement loop

Ricursive’s vision goes beyond faster chip design. The company is building toward what CTO Azalia Mirhoseini calls “rapid AI and hardware co-evolution.” The loop works like this: AI systems design chips, those chips power more capable AI systems, and those AI systems design even better chips. Each generation improves the next.

This isn’t science fiction. Google already demonstrated the concept with AlphaChip and TPUs. Each TPU generation powered the training of AI systems that helped design the next TPU generation. Ricursive is attempting to industrialize and accelerate this process.

“Chips are the fuel for AI, and scaling laws are driving much of the progress,” Mirhoseini explained. “The faster we can make chips that are more custom or better designed for the AIs that we run, the faster we enable this more efficient kind of compute. And that bends the curve for our scaling law.”

What Ricursive is building

While AlphaChip focused primarily on floorplanning—the arrangement of components on a chip—Ricursive is building a full-stack AI platform for semiconductor design. The company aims to automate not just layout, but verification, architecture exploration, and optimization across the entire design flow.

The team is staffed with talent from DeepMind, Anthropic, Apple, and Cadence. The $300 million will fund compute infrastructure expansion and scaling the engineering team—critical for training the AI systems that will design next-generation chips.

The speed of Ricursive’s funding rounds signals investor conviction. Sequoia led the $35 million seed in December 2025. Less than two months later, Lightspeed led the $300 million Series A, bringing total funding to $335 million. For context, most AI startups take 12-18 months between major funding rounds.

The bigger picture

Ricursive sits at the intersection of two critical trends: AI’s insatiable demand for compute and the industry’s shift toward novel chip architectures. As traditional scaling hits physical limits, custom silicon becomes increasingly valuable. Companies like Google, Amazon, and Microsoft are already designing custom AI chips to compete with NVIDIA.

The question is whether AI can design chips better than humans. The AlphaChip results suggest yes—at least for certain tasks. The open question is how much of the 18-36 month design cycle can be automated, and at what quality level. Google’s TPU roadmap demonstrates that AI-assisted chip design already works at scale. Ricursive is betting it can work even better as a standalone company unburdened by corporate priorities.

Guru Chahal, the Lightspeed partner who led the round, called it “the most critical bottleneck facing the AI industry today: the gap between AI advancement and semiconductor capability.” With $335 million in funding and the team that proved AI chip design works, Ricursive has the resources to test that thesis at scale.

If they succeed, chip design shifts from a 2-3 year cycle to weeks. If that happens, the recursive loop kicks in—and AI progress accelerates in ways that are difficult to predict. The founders who taught AI to design chips are now building a company to industrialize that capability. The semiconductor industry is about to find out what that means.