1X Technologies is targeting 50,000 robot shipments in 2027, but the company's biggest hurdle isn't manufacturing capacity,it's keeping those robots reliable once they're deployed in homes and businesses. CEO Bernt Børnich emphasized that the harder task is ensuring robots don't come back for repairs, a shift in focus that reveals how the humanoid robotics industry is moving from lab demonstrations to real-world durability.

What Does 1X's Manufacturing Expansion Actually Look Like?

1X is building out two factories to support its ambitions. The Hayward, California facility is designed to produce about 10,000 units annually at full capacity, though it won't reach that volume until late this year. A second facility in San Carlos will add roughly 100,000 units of annual capacity, with most of that coming online in late 2027. Together, these sites represent 110,000 units of annual production capacity, though the 50,000-unit shipment goal is gated by quality and yield rather than raw output.

What sets 1X apart is its manufacturing speed. Børnich explained that the company can move from a major design change in computer-aided design (CAD) software to a new robot walking off the production line in about four weeks. This rapid iteration stems from 1X's decision to manufacture its distinctive motors and tendon-driven mechanisms in-house, allowing engineering improvements to feed directly back into production without waiting for external suppliers.

How Is 1X Planning to Deploy 50,000 Robots Across Different Markets?

The 50,000-unit shipment target isn't a promise to place 50,000 independent household assistants in homes. Instead, Børnich described a mixed deployment strategy that includes enterprise applications, home deployments, and a developer platform. Some robots will go into structured settings where tasks are more predictable and measurable, while others will be placed with early adopters willing to provide feedback as the technology matures.

1X is also opening its NEO platform to outside developers, expecting a few dominant general-purpose models plus specialized fine-tuned versions, mirroring how the large language model (LLM) ecosystem works. This approach exposes robots to environments and applications that 1X couldn't cover alone, generating diverse real-world experience that feeds back into model improvements.

Why Does Real-World Data Matter More Than Lab Tests?

Børnich framed the manufacturing ramp as serving a second critical purpose: creating a fleet that can learn from attempts, successes, and failures in the physical world. He noted that the company is "diversity-bound rather than data-bound," meaning the challenge isn't collecting more hours of repetitive data but gathering varied experiences across different environments and tasks.

1X's training strategy combines multiple data sources to build more capable models:

• Internet video: Leveraging publicly available footage to understand how humans perform tasks
• Simulation and synthetic data: Creating virtual environments to test robot behavior without physical risk
• Human sensor recordings: Capturing egocentric video and teleoperation data from human operators controlling robots
• Robot learning: Allowing deployed units to learn through their own actions and interactions with the environment

This approach extends the strategy behind 1X's World Model Lab, which trains models to understand how the world changes rather than simply reproduce narrowly demonstrated behaviors. The dependency is demanding: robots need enough capability to generate useful experience, enough safety to attempt unfamiliar actions, and enough deployed units to encounter diverse situations.

What About Safety Before Homes Get Their First Robots?

Børnich acknowledged that home deployment would be uneven and emphasized that safety remains an area of active work. The company hopes to share more formal safety evidence later in the year, though the interview doesn't establish that broad household autonomy or new safety certification has been achieved. Rather than recalling robots for repairs, 1X plans to service early units in the field, allowing the fleet to iterate while customers receive a product worth keeping.

The company is also taking a pragmatic approach to hardware improvements. When problems emerge that are too rare to appear in a hundred robots but could become significant in a fleet of thousands, 1X intends to push those improvements to existing customers through remanufacturing or field service, not just new production units.

When Will General-Purpose Robots Actually Outperform Specialized Ones?

Børnich made a notable concession about the current state of the technology: collecting task-specific data and training a vision-language-action model could today outperform 1X's world model at tasks like laundry folding. However, he argued that such specialization doesn't automatically extend to other activities. He expects specialized models to remain stronger in 2026, but predicts a shift toward more general intelligence in 2027.

He was also dismissive of treating laundry demonstrations as proof of broad capability. While deformable clothing was historically difficult for robotics, folding has become comparatively straightforward with modern AI. Børnich noted that laundry folding wouldn't be 1X's main enterprise application, suggesting the company is focused on tasks with clearer commercial value.

How to Evaluate 1X's Progress on Humanoid Robots

• Manufacturing maturity: Track whether 1X meets its four-week iteration timeline from CAD change to production, a metric that reflects supply chain integration and engineering responsiveness
• Fleet reliability: Monitor how many deployed robots require field service versus recall, the true test of whether the company can keep units working in real homes and businesses
• Data diversity: Assess whether 1X's deployed fleet encounters varied enough environments and tasks to meaningfully improve general-purpose models, not just repeat the same actions
• Safety validation: Watch for formal safety evidence and certifications that would enable broader home deployment beyond early adopters
• Developer ecosystem: Observe how many third-party developers adopt the NEO platform and what applications they build, indicating whether the platform strategy is working

"The real gating item here is really kind of like not just shipping 50,000 units, but ensuring that you don't get them back," said Bernt Børnich.

Bernt Børnich, CEO at 1X Technologies

The distinction between these milestones matters for judging the rollout. A large shipment count, a convincing demonstration of a specific task, and a robot that reliably handles unfamiliar situations are different achievements. Børnich's plan depends on progress across all three, and on early customers providing useful experience while receiving a product worth keeping.

1X's focus on reliability and real-world learning reflects a maturing industry. Rather than chasing headline-grabbing demonstrations, the company is betting that the path to general-purpose humanoids runs through deployed fleets that generate diverse data while maintaining customer trust. Whether that strategy succeeds will depend less on factory capacity and more on whether robots can actually stay in the field without constant repairs.