Robotics in Review: Editors Look Back at 2025

The 2025 robotics landscape was dominated by a surge in humanoid investment, with more than $3.5 billion funneled into startups and a billion earmarked for figure‑AI. This capital influx coincided with breakthroughs in data pipelines and synthetic world models, allowing developers to harvest high‑quality multimodal data without costly tele‑operation rigs. Tools like the Universal Manipulation Interface and AI‑driven simulators such as Google’s Genie turned simulation from a research toy into a core component of production training, setting a new baseline for robot learning efficiency.

Despite the financial firepower, real‑world deployments of humanoid robots remain sparse. Industry leaders highlighted modest milestones—Digit moving 100,000 totes for GXO, and the O2 model assisting with 30,000 BMW tasks—but large‑scale commercial rollouts are still elusive. Safety standards are catching up, with NIST‑ASTM committees drafting guidelines to address the unique risks of human‑like machines. The gap between laboratory performance and field reliability fuels skepticism, as many announced models lack verifiable operational data, especially in the Chinese market where financial disclosures are opaque.

Autonomous vehicle progress mirrored the humanoid narrative: Waymo scaled to dozens of cities, introduced highway driving, and secured international pilots, yet safety incidents underscored lingering challenges. Recent recalls over school‑bus interactions and a high‑profile power‑outage incident in San Francisco revealed weaknesses in edge‑case handling and intersection decision‑making. These events highlight the need for robust end‑to‑end learning safeguards and transparent contingency protocols. As the industry moves toward 2026, the convergence of advanced data engines, emerging standards, and real‑world testing will determine whether robotics and self‑driving cars transition from headline‑grabbing demos to dependable, everyday tools.