Alphabet's robotics software company Intrinsic has open-sourced the foundation of its industrial robotics platform, releasing Intrinsic Core under a permissive Apache 2.0 licence at ROSCon 2026 in Toronto. The move hands the wider robotics community a production-grade control and perception stack that Intrinsic itself depends on ā and reframes the competitive question in industrial automation from who owns the software layer to who builds the best applications on top of it.
Intrinsic, which was folded into Google in February, describes the release as a preconfigured, ROS-compatible environment that runs on local hardware rather than a cloud-tethered service. That detail matters on a factory floor, where latency budgets are measured in milliseconds and network dependence is a liability rather than a feature.
What Is Actually in the Box
The centrepiece is Intrinsic Control, a hardware-agnostic real-time control framework built for fast, sensor-driven behaviour. Its defining capability is mid-trajectory adaptation: a robot can change what it is doing partway through a motion in response to what its sensors report, instead of blindly completing a pre-planned path. Because the framework is hardware-agnostic, developers can swap robot arms, grippers and sensors without rewriting drivers ā historically one of the most expensive forms of lock-in in industrial robotics.
Around that core, Intrinsic has released a set of components that together cover the practical work of getting a robot to manipulate real objects:
- Pose estimation built on NVIDIA's FoundationPose, letting robots locate parts without rigid fixtures or custom jigs.
- Motion planning for collision-free paths through cluttered workcells.
- Grasp planning that adapts gripper strategy to an object's actual orientation.
- Simulation, calibration and ROS drivers to close the loop between virtual testing and physical deployment.
The public repository lists Ubuntu 24.04 LTS or 26.04 LTS as primary operating system prerequisites, with 22.04 also supported, and targets ROS 2 Lyrical Luth. The runtime is a native k3s-based containerised environment, with the core implementation written largely in C++ using Bazel-based tooling ā a stack aimed squarely at teams already comfortable with modern infrastructure practice.
A Reference Design, Not Just Libraries
Alongside the core, Intrinsic published the Open Machine Tending Solution, an open reference design for AI-powered CNC machine tending ā the repetitive work of loading and unloading parts from machine tools. It supports modular customisation for hardware from manufacturers including Universal Robots and FANUC.
Reference designs of this kind often do more practical good than libraries alone. Machine tending is one of the highest-volume, lowest-glamour applications in manufacturing, and the gap between a working demo and a deployable cell is usually filled with integration labour that never gets shared. Publishing a complete, customisable path from arm to finished workflow lowers the cost of the first deployment, which is where most automation projects stall.
Why It Matters
The strategic logic is worth stating plainly. Intrinsic is giving away infrastructure it relies on internally, framing it as a durable commitment to the community rather than a trial balloon. Geoffrey Biggs, chief technology officer of the Open Source Robotics Foundation, called the release both a first step and a demonstration of where robotics development is heading, predicting that platforms like it will make robots considerably easier to produce.
For the industry, three consequences stand out. First, the driver-rewriting tax falls. A hardware-agnostic control layer under a permissive licence means a systems integrator can change vendors without discarding months of work, which shifts negotiating power towards buyers and towards robot makers who compete on hardware quality rather than software lock-in.
Second, it accelerates the physical AI trend that has dominated robotics discussion through 2026. Foundation models for manipulation need a reliable substrate of control, perception and simulation beneath them; without it, every research group rebuilds the same plumbing. Open-sourcing that substrate lets more teams spend their effort on the learning layer.
Third, the choice of Apache 2.0 is deliberate and commercially significant. Unlike copyleft licences, it allows companies to build proprietary products on top without publishing their own modifications, which is the condition most industrial vendors require before adopting outside code at all.
There are caveats. Open-sourcing a core does not open-source Intrinsic's full commercial platform, and the company retains the advantage of having built, tested and operated this code at scale. Adoption will also depend on how actively Intrinsic accepts outside contributions rather than treating the repository as a periodic code drop ā the difference between an open project and merely visible source.
Intrinsic showed the release at ROSCon with live demonstrations, and chief executive Wendy Tan White was scheduled to speak on Wednesday about the company's open source investment, also announcing the top five winners of the AI for Industry Challenge, a competition run with Open Robotics on dexterous cable routing for server manufacturing that was first announced at ROSCon 2025.
