Sift has raised $42M to build the data infrastructure layer for physical AI. As machines generate massive, unstructured sensor data, AI still struggles to interpret it. TheMarketAI Take: the challenge in robotics isn’t just better models but making the physical world legible. Bridging that gap may be key to scaling AI beyond software.

Large language models can train on the internet. Robots cannot. Physical AI must learn from scarce real-world interactions involving vision, touch, motion, and physics. Collecting that data is slow and expensive.

Cisco has invested in World Labs, the spatial-intelligence startup founded by Dr. Fei-Fei Li, underscoring a major shift in AI: from language-based models to systems that understand and operate within 3D environments. World Labs’ Large World Models aim to move AI from “understanding words to understanding worlds.” As spatial and multimodal intelligence emerges, it may define AI’s next decade far more than text-only LLMs.

Physical AI faces hurdles digital AI never had to. Real-world data is scarce, hardware is costly, and edge computing adds complexity. Even with powerful simulators, the “sim-to-real” gap keeps autonomy out of reach — many demos still rely on human “babysitters.” TheMarketAI Take: progress will come from modular systems — robots that master walking, grasping, or vision, not everything at once.