Skip to content
NewsProduct launch

Alphabet's Intrinsic Open-Sources Its Industrial Robotics Platform Under Apache 2.0 at ROSCon 2026

· by Pondero Newsdesk

The short version

Intrinsic released Intrinsic Core on September 22, 2026, publishing the ROS-compatible software stack it uses for real manufacturing deployments under the Apache 2.0 license, alongside an open reference design for CNC machine tending.

Alphabet's Intrinsic Open-Sources Its Industrial Robotics Platform Under Apache 2.0 at ROSCon 2026

For the first time, Alphabet's industrial robotics subsidiary published the software stack it uses for its own manufacturing deployments as open source. Intrinsic released Intrinsic Core on September 22, 2026 at ROSCon 2026 in Toronto, putting a ROS-compatible platform covering real-time control, perception, motion planning, and simulation on GitHub under the Apache 2.0 license.

What Intrinsic Core includes

The release bundles seven capability sets. Intrinsic Control is a hardware-agnostic real-time control framework that lets developers swap robot arms, grippers, and sensors without rewriting drivers. Pose estimation ships with built-in NVIDIA FoundationPose support, delivering six-degrees-of-freedom part detection without custom pipeline wrappers. Motion planning auto-generates collision-free paths. Grasp planning reads live sensor feedback to pick parts dynamically. A Gazebo-powered simulation service runs alongside developers as they build. Camera calibration automates the robot's vision system setup. Pre-configured Intrinsic-ROS drivers connect supported robots, grippers, and 3D cameras.

Per Intrinsic's announcement, all components are ROS-compatible and available on GitHub at intrinsic-ai/intrinsic-core.

Alongside Intrinsic Core, Intrinsic released the Open Machine Tending Solution (OMTS): a reference design for AI-enabled CNC machine tending that runs on Intrinsic Core and supports hardware from FANUC and Universal Robots. Intrinsic described the OMTS as the first in a planned series of open reference solutions.

Robotics teams can skip rebuilding control and calibration code

Robotics software development has historically required each team to build the same foundational pieces from scratch: real-time control loops, camera calibration pipelines, sensor drivers, and motion planners each time a new robot integration begins. Intrinsic Core provides those pieces pre-built and ROS-compatible under a permissive license, meaning developers can start from working production code rather than a blank slate.

The production origin is what separates this from a standalone OSS toolkit. Intrinsic states the release packages the same capabilities it uses day-to-day for real manufacturing deployments, not a parallel project built for the open-source release.

The OMTS reference design targets a concrete market gap. Per the announcement, tens of thousands of fabrication businesses in the US and Europe have little to no automation. A ready-made machine tending reference that supports FANUC and Universal Robots hardware lowers the entry cost for integrators serving that segment.

Intrinsic reported more than 5,000 developers and roboticists across 115 countries entered its first AI for Industry Challenge, per its blog, signaling demand from the developer community for accessible industrial robotics tooling.

What to watch next

The current release covers on-edge inference via NVIDIA FoundationPose but does not include a GPU-accelerated training pipeline. Whether Intrinsic extends Intrinsic Core to cover training is among the near-term questions for developers building full physical-AI pipelines. Whether other robotics platform vendors respond with comparable Apache-licensed stack releases will indicate how broadly this open-source posture spreads in industrial robotics.

Sources