AI robotics has a big data problem. This startup raised $405 million to fix it in surprising ways. (www.businessinsider.com)

FieldAI raised $405 million from heavyweight backers — including Nvidia, Jeff Bezos, Bill Gates, Vinod Khosla, Intel and Samsung — to tackle one of robotics’ core bottlenecks: the lack of large, diverse real-world data for embodied AI. Founded by ex‑JPL engineer Ali Agha and staffed with talent from Google, DeepMind, Waymo, Tesla and Boston Dynamics, FieldAI’s pragmatic strategy is to deploy many modest robots into real environments to perform useful, narrow tasks while continuously collecting sensor and visual data. Those data feed back into models that control a variety of platforms (quadrupeds, humanoids, wheeled robots and passenger‑scale vehicles), creating a deploy→learn→deploy flywheel that accelerates capability growth without needing massive owned assets like Tesla or Amazon. Technically and commercially, the approach matters because it builds the “internet of atoms” datasets that current large models lack. FieldAI is already in construction, energy, manufacturing, urban delivery and inspection across the US, Europe and Japan; one showcased use is automated BIM updates where robots repeatedly photograph and audit sites, improving ROI and enabling compounding use cases (progress tracking, safety checks, inventory). With robots contracted 10× more in H1 2025 than H1 2024, the new funding will productize the stack and scale data collection — a realistic path to improve embodied intelligence models and accelerate practical autonomy in industry.