Google Brings Robotics Startup Intrinsic In-House to Boost AI

Google is pulling its robotics ambitions closer to home. 

The company announced it is bringing Intrinsic — the Alphabet-founded industrial robotics platform often described as the “Android for robotics” — fully under Google to accelerate its push into physical AI.

The move unites Intrinsic with Google DeepMind, Gemini models, and Google Cloud, linking advanced AI research more directly to industrial automation and outlining plans to scale across manufacturing and logistics.

From ‘Other Bet’ to Google’s main stage

Until now, Intrinsic sat in Alphabet’s “Other Bets” portfolio, operating at arm’s length from Google’s core product and infrastructure teams. With this change, it becomes a distinct group inside Google, a structural shift that typically means tighter leadership alignment, closer access to shared resources, and a sharper mandate to ship.

Google is effectively collapsing the distance between robotics development and its mainstream engineering and go-to-market machine. Intrinsic’s platform is now set to plug into Google’s broader enterprise build-and-deploy pipeline, shortening the runway from prototype to production.

When factory robots can see, reason, and react

Physical AI changes how industrial robots operate on the ground.

Traditional robots are programmed for tightly controlled tasks and often require manual resets when something changes. Intrinsic’s approach focuses on systems that can interpret input from cameras and sensors and adjust their movements on the fly.

On an assembly line, that could mean recognizing slight differences in parts before fitting them into place, or correcting alignment without halting production. In logistics settings, it may involve rerouting around obstacles or adjusting to workflow changes without waiting for a technician. The result is fewer stoppages and less reprogramming, with machines better able to handle variation without constant supervision.

Building robotics apps without starting from scratch

A major part of Intrinsic’s pitch is software abstraction. Industrial robotics has traditionally required custom integrations for each robot, sensor, and task, often demanding specialized engineers and long development cycles. Intrinsic is trying to standardize that layer.

Its web-based development environment, Flowstate, lets teams build and test applications in simulation before deploying them to real machines. Developers can assemble prebuilt “skills,” such as modular behaviors for object detection, motion planning, or force control, instead of coding every movement line by line. Those skills can be reused across different hardware setups, whether the system involves robotic arms, cameras, or AI models.

By separating application logic from specific machines, the platform is built to roll out automation across mixed fleets without rebuilding workflows each time. Google says the next step is to expand real deployments and move more systems from simulation to live production, turning research advances into tools that can hold up in day-to-day operations.

Related reading: See how physical AI is transforming manufacturing in the AI-native factory, from factory-floor automation to adaptive robotics workflows.