Particle.news

Big Tech, Startups and Chipmakers Push From Software to Physical Robots

Companies are investing in simulation, high‑fidelity behavior data and new chips to move AI models into real‑world robots over a multi‑year effort.

Overview

  • This week OpenAI posted hires for a new OpenAI Robotics team that it says will build short‑term infrastructure robots and pursue long‑term personal robots using world‑simulation research led by Aditya Ramesh.
  • Chinese startup Jianzhi Robotics announced multi‑hundred‑million‑RMB funding led by Ant Group, Didi and Delian Capital to scale high‑quality, multimodal human behavior data and data‑foundation models for embodied AI.
  • At Computex chipmakers and foundries stressed hardware changes for embodied AI: NVIDIA said its Vera CPU and new PC AI chips target intelligent agents, and TSMC described using NVIDIA GPU libraries to speed lithography, simulation and defect detection in fabs.
  • Automakers and platform companies are reorganizing to prioritize embodied intelligence, with Li Auto creating dedicated embodied‑engineering, embodied‑interaction and embodied‑behavior teams and autonomous driving spun out as its own unit.
  • Consulting and industry commentary temper near‑term expectations, with a Bain survey showing 40% of large firms saw cost cuts of 10% or less from AI and executives warning of data, deployment and chip‑capacity frictions that make commercialization a multi‑year engineering challenge.