Everyone wants to be the layer above the robot

Between August 4 and August 31, 2026, three organisations described a product meant to run on hardware they did not build. NAVER said it would complete its robot ecosystem within the year and open its ARC Brain orchestration platform to third-party cleaning and patrol robots, with humanoids under review as future connected devices. Generalist was valued at $3 billion on a cross-embodiment foundation model intended to drive different bodies from one policy. And ABEJA and Murata Manufacturing demonstrated a dual-arm manipulation policy built on NVIDIA’s Isaac GR00T N1.7.

All three are betting the margin sits above the machine. They are betting on different things transferring.

Three claims about what transfers

Service logic. ARC Brain does not try to teach a robot how to move. It is a cloud control system that coordinates robots with elevators, access gates and building systems, and NAVER’s July launch of ARCBRAIN Flow, a drag-and-drop tool for non-developers, puts the product’s value in composing services — routes, deliveries, notifications — rather than in motor control. “Gathering 10 truly smart robots together does not necessarily make for a good service,” NAVER Labs head Baek Jong-yoon said. “You need a platform above them that oversees the robots.” What has to transfer is an interface, not a skill.

Manipulation policies. GR00T N1.7 is a base model others fine-tune with their own data. What transfers is representation and pretraining, not the finished skill; ABEJA and Murata still had operators teleoperate the dual-arm robot and trained on hundreds of collected demonstrations of their own task. This is the most conservative of the model-based claims and the one with the most working examples behind it.

Embodiment itself. Generalist’s claim is the strongest: one foundation model that works across different robots, with its Gen 1.5 model reported to learn new tasks from video demonstrations of 3 to 12 seconds. If it holds, hardware differentiation collapses. It is also the least demonstrated in public, which is what the valuation is actually pricing.

The wall the model builders run into

Kinematics. A policy learned on one arm’s link lengths, joint limits and dynamics does not move to another for free, and the failure is not graceful — it is a trajectory that was correct on the training body and collides on the target one. That is the same wall sim-to-real transfer has been hitting, arriving from a different direction. It is why the shallow claim (pretraining transfers) has more evidence behind it than the deep one (skills transfer).

There is also an unglamorous version of this business that nobody is valuing at $3 billion: tooling. Evaluation, data management and simulation harnesses are what a fleet operator buys before it buys a policy, and they are indifferent to which model wins.

Why it matters

The layer question determines who keeps the margin as robot unit costs fall. Every hardware maker in the sector is currently arguing that its integration is the differentiator; every software layer is arguing that integration is a transitional cost. Historically the software side of that argument has won when the hardware became interchangeable and lost when it did not, and robot bodies are not interchangeable yet.

NAVER is the odd one out because it walks around the kinematics wall instead of through it. ARC Brain treats each robot as a device that exposes an interface and takes orders, so it does not need a policy to survive a change of body; it already runs services at NAVER’s 1784 headquarters in Seongnam and at Tokyo Midtown Yaesu. The price of that shortcut is that its hardest dependency is commercial, not technical: other manufacturers have to agree to be orchestrated, and NAVER has not named which ones have. That is the most deployable claim on the platform position and the least defensible one, and the next thing to watch is a named third-party robot maker on ARC Brain.

Follow-up coverage

Later reporting on this story: FieldAI signs term sheet for $700M at a reported $10bn valuation (Oct 7, 2026).