Gemini models integrated into Boston Dynamics’ Spot to add multimodal perception

Boston Dynamics has started using Google’s Gemini models inside the software that powers Spot inspection jobs in factories and industrial sites. (bostondynamics.com) The company said Orbit AIVI-Learning now integrates Gemini and Gemini Robotics-ER 1.6 through a partnership with Google Cloud and Google DeepMind. Spot robots use Orbit to patrol facilities, collect images and flag anomalies. (bostondynamics.com) (deepmind.google) A visual-language model works like a captioning system plus a reasoning engine: it looks at an image, connects it to text, and answers questions about what it sees. Google said Gemini Robotics-ER 1.6 is built for “embodied reasoning,” meaning it helps robots interpret the physical world before another system takes action. (deepmind.google) (blog.google) In Boston Dynamics’ example, Spot can read circular pressure gauges, vertical level indicators and digital displays during inspection rounds. Google said instrument reading emerged from work with Boston Dynamics on industrial inspection use cases. (deepmind.google) (bostondynamics.com) Boston Dynamics said the system can also connect camera feeds to facility manuals, maintenance history and operator questions in natural language. The company framed that as a way to reduce custom rule-writing for each site. (bostondynamics.com) The Spot news is narrower than a full onboard autonomy upgrade. Boston Dynamics’ posts describe Gemini as part of Orbit AIVI-Learning and a demo using Gemini Robotics-ER 1.5 for embodied reasoning, not a blanket replacement for Spot’s existing navigation and control stack. (bostondynamics.com 1) (bostondynamics.com 2) Google and Boston Dynamics had already announced a separate January 5, 2026 partnership focused on Atlas, Boston Dynamics’ humanoid robot. That deal said joint research using Gemini Robotics foundation models and a new Atlas fleet would begin in 2026. (bostondynamics.com) Google is also pushing Gemini into engineering software. Cadence said on April 15 that it is optimizing its ChipStack AI Super Agent with Gemini on Google Cloud for chip design and verification workflows. (cadence.com) Cadence said the system delivers up to 10x productivity improvements across digital design, testbench development, verification planning, regression management and automated debug. Cadence launched ChipStack in February and called it the first agentic workflow for automating chip design and verification. (cadence.com 1) (cadence.com 2) The common thread is that Gemini is being used less as a chatbot and more as a perception-and-reasoning layer inside existing tools. In Boston Dynamics’ case that means reading the physical world; in Cadence’s case it means navigating the logic of chip-development tasks. (deepmind.google) (cadence.com) For now, both companies are selling augmentation rather than full automation. The pitch is that Gemini helps robots see and helps engineers move faster, while the underlying control systems and design tools stay in place. (bostondynamics.com) (cadence.com)