We are thrilled to announce Radical Ventures’ investment in Ricursive Intelligence‘s $300M Series A round as the company seeks to redefine the relationship between AI software and hardware.
For decades, AI researchers have pursued systems that can improve themselves without human intervention. Ricursive Intelligence is turning this vision into reality by going all the way to the bottom of the technology stack and starting with semiconductor design.
Ricursive is developing reinforcement learning systems that automate chip design, compressing the time required to design a new chip from a few years to a few weeks. As AI designs better chips, those chips enable more powerful AI, which in turn designs even better chips, creating a recursive loop that rapidly compounds.
Today’s AI revolution is bottlenecked by specialized hardware. Custom silicon could unlock transformative performance gains, but chip design is too slow and expensive. Ricursive changes this by enabling true hardware-software co-evolution, where models and chips can be developed in parallel, purpose-built for one another.
Anna Goldie and Azalia Mirhoseini are literally the best people on planet Earth to build this company. They co-led Google DeepMind’s groundbreaking AlphaChip project, which used reinforcement learning to automate chip layout and has been deployed in production across multiple generations of TPUs, smartphones, and other Google hardware. They co-founded Google’s ML for Systems team and were among the first 20 employees at Anthropic, working closely with Dario Amodei and the founding team.
With an unparalleled team and a wildly ambitious vision, Ricursive has the potential to become one of the defining companies of the AI era.
We are honored to support Anna, Azalia, and the entire Ricursive team on their journey ahead.
AI News This Week
-
Waabi Raises $1B and Expands Into Robotaxis with Uber (TechCrunch)
Radical Ventures portfolio company Waabi secured $1 billion in funding to expand its autonomous driving technology from trucking into robotaxis. The Toronto-based company closed an oversubscribed $750 million Series C round led by Khosla Ventures and G2 Venture Partners, with participation from Radical Ventures, NVentures (Nvidia’s venture arm) among others. Waabi also received $250 million in milestone-based investment from Uber (already an existing investor) to launch at least 25,000 robotaxis exclusively on its platform. Waabi’s technology doesn’t require extensive mapping or prior observation of scenarios to learn new environments, potentially reducing validation time. The company previously partnered with Volvo to produce self-driving trucks.
-
Moltbook, a Social Network Where AI Agents Hang Together, May Be ‘The Most Interesting Place on the Internet Right Now’ (Fortune)
Moltbook is a social network built for AI agents, where bots post, interact, and share information much like humans do on social networks. It emerged alongside Open Claw (formerly Clawdbot), the viral open-source AI personal assistant that connects to messaging apps and automates tasks on a user’s behalf. With over 1.5 million agents currently active on Moltbook, LLM agents discuss technical topics, share automation strategies and comment on their relationship to humans. The platform and tool pose cybersecurity concerns, including the risk that agent-to-agent communication could create a new channel for data leaks. As AI research leader Andrej Karpathy warns, “what we are getting is a complete mess of a computer security nightmare at scale.”
-
A New Test for AI Labs: Are You Even Trying to Make Money? (TechCrunch)
The AI industry faces a unique challenge: determining which foundation model labs are pursuing commercial success versus pure research. Radical Ventures portfolio company World Labs exemplifies rapid commercial progress, shipping both a world-generating model and a commercial product within 14 months after coming out of stealth.
-
Researchers Are Using A.I. to Decode the Human Genome (NYT)
Google DeepMind unveiled AlphaGenome, building on AlphaFold’s protein structure breakthrough to tackle DNA sequences. Trained on three billion base pairs and thousands of gene activity experiments, the model predicts how mutations affect gene regulation across 11 biological processes. The system represents state-of-the-art performance in mutation analysis, demonstrating how AI can uncover connections between distant genetic changes and disease outcomes that took researchers years to validate experimentally.
-
Research: A.I. and Our Economic Future (Stanford/NBER)
Stanford economist Charles Jones argues AI will likely be “the most important technology we have ever developed,” with broader economic effects than electricity or semiconductors. He uses task-based economic models to explore two scenarios: AI maintaining historical 2% GDP growth or accelerating growth rates to 5% annually if AI expands continuously into new applications. Jones expects AI’s cumulative impact to exceed the internet’s by 10x over the next half-century.
Radical Reads is edited by Ebin Tomy (Analyst, Radical Ventures)
