Figure and 1X Ramp Humanoid Production Before Demand Exists
Figure is now building 55 humanoid robots per week — with commercial use cases still listed as "in development." Meanwhile, 1X just opened a 58,000 sq ft factory in Hayward, CA, declaring the era of "abundant home robots" has arrived. The hardware is scaling. The killer app is not.
Explanation
Two humanoid robot companies made major manufacturing announcements this week, and both carry the same awkward subtext: production is outpacing purpose.
Figure says it can now produce 55 robots per week. The output is earmarked for internal R&D, data collection, end-to-end housework experiments, and "commercial use-case development" — that last phrase doing a lot of heavy lifting. Building robots at scale to figure out what they're for is an expensive way to run a research program, but it does generate training data fast.
1X went further. The company opened its NEO Factory in Hayward, California — 58,000 square feet, 200-plus employees, fully vertically integrated. Motors, batteries, transmissions, sensors, structures, and final assembly all under one roof. Consumer shipments are planned for 2026. The press release calls it "the critical milestone that turns the vision of abundant, general-purpose home robots into reality." IEEE Spectrum's editors appended a dry "Scale will fix everything...?" — which is doing the journalistic work here.
The vertical integration story is genuinely interesting. Building every critical component in-house shortens the iteration loop and avoids supply chain exposure — lessons learned watching Tesla and, more recently, watching Chinese robotics firms move fast. Whether it translates to a robot that can reliably fold laundry is a separate question.
The broader context from this week's robotics roundup: Agility's Digit is being pushed to balance on one leg to stress-test sim-to-real pipelines; researchers are bolting distributed tactile sensors onto humanoids for contact-rich manipulation; and Unitree G1s from China are already sitting in OpenAI and Nvidia labs. The race is real. The finish line is still being drawn.
Watch whether Figure's 55-per-week output translates into any announced commercial contracts by end of 2025 — that's the number that would change the picture.
Figure's 55-units-per-week cadence is notable less for the absolute number than for what it signals about strategy: high-volume hardware production as a data flywheel. Each robot in an R&D or housework-trial setting generates proprioceptive, visual, and task-completion data that feeds policy training. If you believe data is the primary bottleneck to general manipulation — not actuator quality or compute — then overbuilding hardware ahead of commercial demand is rational. The risk is burn rate; 55 robots/week at even a conservative $50K/unit cost-to-produce is ~$143M/year in hardware alone before any revenue.
1X's NEO Factory announcement leans hard on vertical integration as a differentiator. The claim — motors, batteries, transmissions, sensors, structures, and final assembly in-house across 58,000 sq ft with 200+ staff — mirrors the playbook that gave Tesla margin advantages over traditional OEMs and that Chinese firms like Unitree have used to undercut Western competitors on price. The "true American scale" framing is pointed: it's a geopolitical pitch as much as a manufacturing one, arriving at a moment when Unitree G1s are already shipping to U.S. research labs via domestic resellers.
The technical signals elsewhere in the roundup are worth tracking. Agility's one-leg balance demo for Digit is explicitly a sim-to-real stress test — the team notes that "even the slightest model mismatches can lead to instability," which is an honest admission of how fragile current locomotion policies remain under distribution shift. The Humanoid Touch Dream system (VR teleoperation + reinforcement-learned lower body + distributed tactile sensing + multimodal transformer policy) represents the current state-of-the-art for contact-rich manipulation — and it still requires a human in the loop for data collection.
The open question neither company addresses: what is the actual error rate on unstructured household tasks, and how does it scale with robot count? Producing 55 units/week into internal trials is only valuable if the feedback loop from those trials closes fast enough to matter. Consumer shipments in 2026 is a hard claim from 1X — watch for any slip in that timeline as the real signal on readiness.
Reality meter
Why this score?
Trust Layer Figure and 1X are scaling humanoid robot manufacturing to significant weekly/annual volumes before validated commercial use cases exist, betting that hardware scale will accelerate the path to deployment.
Figure and 1X are scaling humanoid robot manufacturing to significant weekly/annual volumes before validated commercial use cases exist, betting that hardware scale will accelerate the path to deployment.
- Figure has reached a production rate of 55 robots per week, allocated to internal R&D, data collection, housework trials, and commercial use-case development.
- 1X's NEO Factory in Hayward, CA spans 58,000 square feet and employs over 200 team members, with full vertical integration covering motors, batteries, transmissions, sensors, structures, and final assembly.
- 1X states consumer shipments of NEO are planned for 2026 and describes the factory opening as 'the critical milestone that turns the vision of abundant, general-purpose home robots into reality.'
- Unitree G1 humanoid robots (Chinese-made) are already present at OpenAI, Nvidia, and top academic institutions in the U.S., shipped via a Long Island-based reseller.
- Figure's own description lists 'commercial use-case development' as still ongoing — production is scaling ahead of proven demand, which IEEE Spectrum's editors explicitly flag as odd.
- 1X's announcement is self-issued marketing copy with no third-party validation of production volumes, unit costs, or task-performance benchmarks.
- No commercial contract numbers, pricing, or performance metrics are disclosed by either company, making the 'ready for deployment' narrative unverifiable from the source.
Production milestones (55 units/week, factory opening) are concrete and sourced, but commercial readiness is explicitly unproven by the companies' own language — warranting a moderate reality score.
1X's press release language ('turns vision into reality,' 'true American scale') is maximalist and unsupported by performance data; Figure's framing is more measured but the gap between production rate and use-case maturity is real.
If either company achieves reliable general-purpose home robot deployment in 2026, the impact is high; but the source provides no evidence that the capability gap has been closed, keeping near-term impact moderate.
- 1 source on file
- Avg trust 40/100
- Trust 40/100
Time horizon
Community read
Glossary
- proprioceptive data
- Information about a robot's own body position, movement, and internal state—such as joint angles and forces—collected during operation to improve how it learns to move and interact with objects.
- policy training
- The machine learning process of teaching a robot's decision-making system (its 'policy') how to perform tasks by learning from collected data and experience.
- vertical integration
- A manufacturing strategy where a company controls multiple stages of production—from raw materials and components to final assembly—rather than outsourcing to suppliers.
- sim-to-real
- A technique where robots are first trained in computer simulations and then tested in the real physical world to verify that the learned behaviors actually work despite differences between simulation and reality.
- distribution shift
- A situation where a robot encounters real-world conditions or tasks that differ significantly from the training data it learned from, potentially causing it to fail or perform poorly.
- contact-rich manipulation
- Robotic tasks that require precise sensing and control of physical contact with objects, such as grasping, pushing, or assembling items where touch feedback is critical.
What's your read?
Your read shapes future topic weighting.
Your vote feeds topic weights, community direction and future prioritisation. Open community direction
Sources
Optional Submit a prediction Optional: add your prediction on the core question if you like.
Prediction
Will either Figure or 1X announce a signed commercial deployment contract for 100+ humanoid robots before the end of 2025?