Alphabet's Intrinsic Open-Sourced Its Industrial Robotics Base Under Apache 2.0, and the Repo Says It Is Not an Officially Supported Google Product

Intrinsic, the robotics software company owned by Alphabet, announced Intrinsic Core at ROSCon 2026 in Toronto on September 22, with the code landing on GitHub the same day under the permissive Apache 2.0 license. Founded in 2021, Intrinsic was incorporated into Google's operations in February to accelerate work on physical AI. Intrinsic Core provides a local runtime, SDK and hardware-agnostic real-time control framework for industrial robotics, running on local hardware through a pre-configured software environment. Its components include Intrinsic Control, a hardware-agnostic real-time control framework with sensor-based control that lets robots adapt behavior mid-trajectory, so developers can swap arms, grippers and sensors without rewriting drivers; intrinsic_perception, which standardizes camera and point-cloud interfaces with built-in NVIDIA FoundationPose support for 6-DoF pose estimation, enabling manipulation of un-fixtured parts; plus motion and grasp planning, simulation services powered by Gazebo, camera calibration and preconfigured ROS drivers, with ROS 2 interoperability bridging C++/Bazel development to the wider ROS ecosystem. Released alongside it is the Open Machine Tending Solution, an open reference design for CNC machine tending that runs on both Intrinsic Core and the Open Robotics Suite and can be customized for Fanuc and Universal Robots hardware. Intrinsic says robotics teams currently spend hundreds of hours coding capabilities or infrastructure from scratch, and targets shops in the US and Europe, where it says only 8% use any automation. Brian Gerkey called the ROS ecosystem "very strong" while noting that "doesn't mean it's complete or perfect." The GitHub repository carries a disclaimer that this is not an officially supported Google product.

Two Companies Moved in Opposite Directions on Openness in the Same Week

On September 21 this site covered Qwen-Image-2.1: the weights shipped as expected, but the license stepped back from Apache 2.0 to research-only, with commercial use requiring a separate negotiation. Today's story runs the other way — an Alphabet subsidiary releasing an entire industrial robotics base, under Apache 2.0 exactly. Two days apart, opposite directions, same underlying logic: **open source is not a position, it is a choice each company makes based on where it sits in the value chain.** If the model is your product, you pull back. If your business is elsewhere and wider adoption of the base helps you, you open up. Intrinsic belongs to the second category. It does not make money selling a control framework; it wants more production lines running on the abstractions it defines. When judging any act of "open sourcing," asking what the company makes money from gets you to an answer faster than reading the license — and then you go read the license to confirm.

That Disclaimer Carries Real Weight on a Factory Floor

The repository says plainly: this is not an officially supported Google product. For an individual developer that line barely matters. For a factory it decides whether this is prototyping material or production material — **who do you call when the line stops.** Apache 2.0 grants you the right to maintain it yourself, and equally the obligation to. With no support commitment, triage, patching and tracking upstream changes all land on you, or on an integrator you pay to carry them. That is not a reason to avoid it. It is a reason to price "who backstops this" as a separate line item when evaluating it. For shops that already have an automation team, the math is favorable. For shops that do not, open source saves the license fee; it does not save the headcount.

The 8% Figure Explains the Whole Commercial Logic

Intrinsic says it targets shops in the US and Europe, where only 8% use any automation. That sentence makes the intent clear: **this is not about taking share inside the ROS ecosystem, it is about growing the pie.** What that 92% lacks was never algorithms. It lacks the several hundred hours it takes to assemble something that runs, and an abstraction layer that does not require rewriting drivers for every machine. So the two most substantial pieces of this release are also the least glamorous. First, swapping arms, grippers and sensors without rewriting drivers. Second, the Open Machine Tending Solution — a reference design for one specific operation, CNC machine tending, already adapted for Fanuc and Universal Robots hardware. Smaller shops typically run mixed-brand equipment accumulated over years, so unifying the driver layer matters more to them than any model capability; and a reference design you can modify removes the most expensive question of all, which is where to start. Also worth recording: intrinsic_perception ships with NVIDIA FoundationPose support for 6-DoF pose estimation, and what that buys is **manipulating parts with no locating fixtures.** On a real floor that means one less batch of tooling and one less line-layout change — the kind of saving that decides whether a project gets approved, far more often than a few percent of cycle time does.

The Software Side Just Had an Incident About Exactly This Layer

Set this alongside recent stories here and you see one strategy playing out on two fronts. Plugin4Shell, covered on September 21, was about a distribution channel — plugin marketplaces pushing capability onto millions of developer machines — and what happens when that channel fails: four products compromised at once. Today's story is about a runtime and SDK pushing capability onto production lines. **Owning the base and the abstraction layer, so others build against interfaces you define, is the same play.** And because it is the same play, the lesson from the software side applies here. Once many production lines depend on one open-source runtime, a supply chain problem becomes everyone's problem — and the line about this not being an officially supported product means there is no committed upstream response time. It is very early; Intrinsic Core only just landed. But if adoption does get broad, the question worth asking ahead of time is who does the security maintenance, and on what cadence patches ship.

via: Intrinsic's blog post, the intrinsic-ai/intrinsic-core repository, The Robot Report, SiliconANGLE