Boston Dynamics just plugged Google DeepMind’s brain into its robot dog. And the results are the clearest signal yet that physical AI is about to get very real, very fast.

The company announced that Spot, its quadruped inspection robot, is now equipped with Gemini Robotics-ER 1.6, a high-level embodied reasoning model from Google DeepMind. In plain English: Spot can now think before it acts. It can read analog thermometers and pressure gauges. It can reason about what it sees. And it can detect problems in industrial environments without a human telling it what to look for.

That last part is the breakthrough. Previous industrial robots could follow scripted routines: go here, take a photo, upload it. Spot with Gemini can walk through a factory, notice a gauge reading that looks abnormal, reason about whether it matters, and flag it. Autonomously. That is not automation. That is judgment.

Why This Partnership Changes the Game

Boston Dynamics has the best robot hardware in the Western world. Google DeepMind has the best reasoning models. Put them together and you get something neither could build alone: a robot that moves like an athlete and thinks like an engineer.

The Gemini Robotics-ER 1.6 model adds spatial reasoning, instrument reading, and what Google calls autonomous task detection. According to IEEE Spectrum, one new capability discovered through the Boston Dynamics collaboration is instrument reading: the ability for robots to accurately parse complex analog gauges and sight glasses. The kind of thing that currently requires a trained human technician walking around with a clipboard.

Google is also opening this up to everyone. Gemini Robotics-ER 1.6 is available today via the Gemini API and Google AI Studio. That means any robotics startup can give their hardware the same reasoning brain that Spot just got. The implications for the physical AI market are enormous.

The Race Between Chatbots and Robots

While OpenAI and Anthropic fight over who has the best chatbot, Google is quietly building the infrastructure for physical AI. Gemma 4 runs on phones. Gemini runs in robots. Google Cloud provides the compute. It is a full-stack physical AI strategy, and nobody else has one.

The timing here matters. China deployed humanoid robots in live factory shifts last week. Japan is using robots to fill labor shortages. South Korea’s humanoid robot market is projected to hit $583 million by 2030. The West’s response cannot just be better chatbots. It needs robots that can think. Boston Dynamics and Google just showed they can build one.

Google also claims Gemini Robotics-ER 1.6 is its safest robotics model to date, with superior compliance with safety policies on adversarial spatial reasoning tasks. Given that these robots are walking around factories with heavy machinery, that claim better hold up. Because if this works, every oil refinery, power plant, and manufacturing floor in the world is about to get a robot inspector that never sleeps, never misreads a gauge, and never calls in sick.