Fei-Fei Li’s World Labs Secures $1 Billion, Ricursive Gains $335 Million, and Google’s Gemini Adds Lyria 3

Recent investment activity signals continued momentum in the artificial intelligence sector, with World Labs, founded by Fei-Fei Li, securing a substantial $1 billion investment, Ricursive receiving $335 million for AI chip design, and Google integrating its Lyria 3 music generation model into Gemini. These developments highlight the growing focus on both foundational AI models and the specialized hardware required to power them.

World Labs’ $1 Billion Funding Round

Fei-Fei Li’s World Labs has secured $1 billion in funding, building on a previous $230 million raise in September 2023 and a $200 million investment from Autodesk in November 2025. [1] The company is focused on developing “world foundational models” (WFMs), which simulate real-world environments to train AI agents. These models are crucial for advancing robotics and enabling AI to interact more effectively with the physical world.

WFMs are neural networks that predict outcomes based on text, image, or video input, and are being developed by companies including Google, Nvidia, Meta, Physical AI, and Skild AI. [3]

Ricursive’s $335 Million for AI Chip Design

Ricursive, an AI chip design company, has raised $335 million to develop specialized hardware for AI workloads. This funding underscores the increasing demand for chips optimized for the unique computational requirements of AI models. The company aims to address the growing need for more efficient and powerful hardware to support the rapid advancements in AI.

Google’s Gemini Integrates Lyria 3 for Music Generation

Google has expanded the capabilities of its Gemini large language model by integrating Lyria 3, its music generation model. [1] This integration allows Gemini to create original music based on user prompts, demonstrating the growing trend of multimodal AI – systems capable of processing and generating multiple types of content, including text, images, and audio.

DeepMind’s SIMA 2 Agent

Google DeepMind has previewed SIMA 2, the next generation of its generalist AI agent. SIMA 2 leverages the Gemini language model to understand and interact with virtual environments. [1] SIMA 2 builds upon the foundation laid by SIMA 1, which was trained on video game data to learn how to play 3D games. SIMA 2 doubles the performance of its predecessor and is capable of self-improvement based on its own experience.

The Intersection of AI and Robotics

These developments collectively point to a growing convergence between AI and robotics. World foundational models are essential for training robots to navigate and interact with the real world, while specialized AI chips provide the necessary computational power. The integration of AI models like Lyria 3 into platforms like Gemini demonstrates the potential for AI to generate creative content and enhance user experiences.

Looking Ahead

The continued investment in AI infrastructure and foundational models suggests that the pace of innovation in this field will likely accelerate. Further advancements in WFMs, AI chip design, and multimodal AI are expected to drive progress in robotics, automation, and a wide range of other applications. The collaboration between companies like World Labs and Autodesk, and the ongoing research at Google DeepMind, highlight the importance of partnerships and interdisciplinary approaches in unlocking the full potential of AI.