Ricursive Intelligence
Ricursive Intelligence is a frontier AI lab building self-improving systems, starting with AI-driven semiconductor design. Its goal is to close the feedback loop in which AI designs better silicon, and that silicon enables more capable AI, accelerating the path toward artificial superintelligence.
At a Glance
- Semiconductor companies and chip designers
- Hyperscalers and cloud providers
- AI labs and companies building custom AI accelerators
- GPU, CPU, and other processor manufacturers
- +1 more
AI Tools by Ricursive Intelligence
(1)
Ripple-65B Preview
Open Weight Agentic Coding LLM
Discussions
No discussions yet
Be the first to start a discussion about Ricursive Intelligence
Latest News
Ricursive releases Ripple-65B Preview, a 65B open-weight model with up to 1M context and reported 80.52 Terminal-Bench 2.1 score.
Anna Goldie and Azalia Mirhoseini named to TIME100 AI 2026; TIME profiles Ricursive's AI-for-chip-design strategy.
TechCrunch reports how Ricursive raised $335 million at a $4 billion valuation in four months and details its target customers and AI chip-design approach.
Ricursive announces a $300 million Series A at a $4 billion valuation, led by Lightspeed Venture Partners.
Products & Services
A full-stack platform using AI and distributed computing to automate and optimize stages of semiconductor design, with the longer-term aim of making chip design iterative and tightly co-designed with AI models.
An Apache 2.0 open-weight, 65-billion-parameter dense Transformer general-purpose model and public variant of the proprietary model used by Ricursive's chip-design platform. It was pretrained from scratch on 12 trillion tokens, supports up to 1,048,576 tokens of context, and is optimized for reasoning, coding, and agentic workloads.
Market Position
Ricursive positions itself as a full-stack AI-for-chip-design platform rather than a chip manufacturer or Nvidia competitor. Its recursive co-design thesis and AlphaChip-derived expertise differentiate it from traditional EDA incumbents such as Synopsys and Cadence, while its intended customers include Nvidia, AMD, Intel, hyperscalers, and other chipmakers. Ripple-65B is positioned as a relatively small, efficient open-weight model competing on agentic coding and reasoning against much larger open-weight models.
Leadership
Founders
Anna Goldie
Co-founder and CEO. Goldie co-founded and led the ML for Systems team at Google Brain, co-led the AlphaChip project, was an early employee at Anthropic working on Claude, and was a Senior Staff Research Scientist at Google DeepMind working on Gemini. She holds a Stanford PhD in NLP and three MIT degrees.
Azalia Mirhoseini
Co-founder and CTO. Mirhoseini co-led AlphaChip with Goldie at Google, where she was a Senior Research Scientist at Google DeepMind; she is also a Stanford computer science professor and holds a PhD from Rice University.
Executive Team
Anna Goldie
Co-founder and Chief Executive Officer
Former Google Brain ML-for-Systems leader and AlphaChip co-lead; early Anthropic employee; former Senior Staff Research Scientist at Google DeepMind working on Gemini.
Azalia Mirhoseini
Co-founder and Chief Technology Officer
Former Google/Google DeepMind research scientist and AlphaChip co-lead; Stanford computer science professor; researcher in AI-driven chip design and machine learning systems.
Founding Story
Goldie and Mirhoseini had independently been pursuing the idea of using AI for chip design and later co-created AlphaChip at Google. They started Ricursive because semiconductor design was becoming a primary bottleneck for AI: their initial vision was to use AI and distributed computing to compress the chip-design cycle, then create a recursive loop in which improved chips train improved AI models.
Business Model
Revenue Model
The semiconductor platform is positioned as a commercial full-stack software/platform offering for chipmakers and hyperscalers, but Ricursive has not publicly disclosed its enterprise pricing. Ripple-65B is distributed under Apache 2.0 and is monetized for API users by third-party inference providers.
Pricing Tiers
OpenRouter-hosted API pricing; the model is also available through Together AI, Fireworks, GMI Cloud, and Hugging Face.
Weights are available on Hugging Face; infrastructure and compute costs are borne by the user.
Target Markets
- Semiconductor companies and chip designers
- Hyperscalers and cloud providers
- AI labs and companies building custom AI accelerators
- GPU, CPU, and other processor manufacturers
- Developers and researchers needing open-weight coding and reasoning models
- Automated AI accelerator and custom semiconductor design
- Chip architecture, RTL, verification, and physical-design workflows
- Co-designing AI models and the hardware that runs them
- Reducing chip-design time, cost, and energy while improving performance
- Agentic coding and reasoning with Ripple-65B
- Local or hosted inference of an open-weight language model