1X Technologies CEO on deploying 100 NEO humanoid robots into homes to build the data loop that scales

Summary

1X Technologies is targeting a consumer home launch for its NEO humanoid robot by end of 2025, with roughly 100 units produced across multiple hardware revisions to date. The company, founded 10 years ago by CEO Bernt Øivind Børnich, is currently running in-home testing among employees and preparing early NDA customer deployments ahead of a broader market release.

The consumer home focus is a deliberate data strategy, not a product positioning choice. Factory environments plateau at roughly 20 to 30 hours of useful training data before yielding no further learning gains. Homes, by contrast, generate continuous variance through social context, novel layouts, and unpredictable daily activity — exactly the diversity that drives model intelligence at scale.

The core data loop 1X is building works like this: deploy robots into homes at partial capability, perhaps a 50% task success rate, then use real-world feedback — positive reinforcement when the robot succeeds, correction signals when it fails — to compound learning over time. Teleoperation and simulation are treated as bootstrapping tools only. Neither scales: simulation lacks the physical fidelity needed for complex manipulation tasks like peeling a shrimp, and teleoperation is too labor-intensive to sustain at volume.

On AI architecture, Børnich is direct: the path to generalized robot intelligence is a large transformer model trained on enormous volumes of action-outcome data, not just observational internet data. The key distinction is capturing what the robot planned, what it did, and what resulted — a feedback loop that static text, image, or video datasets cannot replicate.

Safety is a hard constraint on the data strategy itself. For robots learning through real-world trial and error inside homes, hallucinations and physical errors cannot be tolerated at the rate acceptable in software. A robot misplacing knives from a dishwasher in a home with children is not a recoverable product experience. 1X's engineering philosophy — lightweight hardware, no exotic materials, tolerances closer to a refrigerator than a car — is designed specifically to make autonomous in-home operation safe enough to actually run.

On supply chain, 1X positions itself as a deep vertical integrator. The company manufactures its own components, is pursuing patents on a novel aluminum die-casting process, and is currently building a US factory. Børnich acknowledges that sourcing copper, aluminum, and steel domestically is significantly more expensive than in China, and that US manufacturing lacks the co-located industrial zones that make Chinese production efficient. His view is that the more durable competitive gap is not in materials but in manufacturing knowledge, which takes far longer to develop domestically than physical supply chain infrastructure. He frames onshoring as a generational workforce challenge as much as a logistics one.

1X's longer-term thesis is that in-home deployment is the shortest path to what Børnich describes as an abundance of artificial labor — robots building robots, constructing data centers, managing energy infrastructure, and running chip manufacturing — a transition he believes is years, not decades, away.