The 1X NEO bet on consumer humanoids
Three production humanoids in 2026, none existed a year ago. 1X is going after the hardest market segment first — the home — with transparent pricing and a confirmed delivery window. Here's the bet, and the unsolved problem.
The map of the market
In a twelve-month window we now have three meaningfully different humanoid programs in production, each going after a different segment:
- Boston Dynamics Atlas — electric, 1.9m, 90kg, 56 degrees of freedom, deploying first at Hyundai's Robotics Metaplant Application Center. Industrial. Paired with Google DeepMind on the brain side.
- Figure 03 — already piloting in BMW's Spartanburg plant, optimized for high-volume manufacturing tasks.
- 1X NEO — pre-orders open this month, positioned as the first consumer-ready home humanoid, with transparent pricing and a confirmed 2026 delivery timeline.
1X is the outlier. The other two are betting on industrial deployments where the customer is a factory with structured tasks, predictable environments, and a procurement department that can absorb a five-year payback cycle. 1X is betting on the kitchen.
Why the consumer segment is hardest
The factory humanoid problem is well-defined: same task, same lighting, same coworkers, same environment, every day for ten years. Whatever the policy gets wrong gets retrained out of it in simulation.
The home humanoid problem is the opposite of every one of those properties. The lighting changes hourly. The "coworkers" include children, pets, and elderly relatives. The environment changes every time someone moves a chair. The set of possible tasks isn't finite — it's whatever the household happens to need this week.
And the safety bar is also different in kind, not just in degree. A factory humanoid that injures a worker triggers a labor lawsuit. A home humanoid that injures a child triggers the end of the entire product category.
If 1X gets this right at scale, they've solved the harder version of the problem. If they get it wrong even once at meaningful publicity scale, they may also have solved it for everyone else — by killing the category for a decade.
The pricing transparency is doing real work
Industrial humanoids hide their list price. Atlas, Optimus, Figure 03 — you don't see published per-unit pricing because the customer is buying a system bundle: hardware, integration, service contract, retraining for new tasks. The "price" is whatever the enterprise sales motion settles on with each buyer.
Consumer hardware can't do that. You have to publish a number. If the number is too high, you have no customers. If it's too low, you can't manufacture profitably.
The fact that 1X opened pre-orders with public pricing and a confirmed delivery window is its own signal. Either they have line-of-sight to unit economics that work at the announced price, or they're willing to subsidize early units to seed the category. Both are reasonable bets at this stage. Neither is something the industrial humanoid players are willing to do yet.
The unsolved problem nobody has admitted
Embodied foundation models are not where the language models are. The same scale-and-data flywheel that produced GPT-5.5 doesn't have an obvious analogue in robotics, because actuation experience is much harder to collect than text. Google DeepMind's partnership with Boston Dynamics is partly an admission of this — combining the hardware shop with the foundation-model shop because neither has the full stack alone.
1X's bet on the home requires them to either:
- Have a quietly-class-leading embodied policy in-house (possible — they've been working at it since before the OpenAI investment), or
- Plan to ship a product that's narrower than "general home assistant" — closer to "folds laundry, loads dishwashers, fetches things from upstairs," with a strong human override layer.
The second strategy is the realistic 2026 product. The first is the 2030 narrative. The interesting question is which one the early NEO units actually deliver.
What to watch
Three signals through 2026 H2:
- Whether the early deliveries include any video of unstructured task performance. A controlled-environment demo is meaningless at this point — every program has those. A first-time-in-a-real-home video is the only meaningful proof.
- How NEO handles the failure case. Every home humanoid will have a "the robot did something wrong" incident. The interesting variable is how 1X — and the customers — respond. Quietly, with a software update? Or loudly, with a product change?
- Whether other consumer humanoids enter the market in 2026. The factory players (Atlas, Figure) don't have to follow 1X home. If they do — if Atlas or Figure announces a consumer variant — then 1X has validated the segment for everyone. If they don't, 1X has the segment to itself for as long as they can hold it.
The honest read
The industrial humanoid bet has obvious math. You replace a $50k/yr worker with a $100k machine that runs three shifts, and the ROI is straightforward.
The home humanoid bet doesn't have that math. Homes don't run three shifts. The value has to come from time saved on annoyance work, and the dollar value of that is much harder to pin down. But it's also much larger if it works — there are vastly more homes than there are factories, and the household labor market is enormous in aggregate even if it's invisible.
1X is betting on the larger and harder market. The other two are betting on the smaller and easier one. The first humanoid program to reach 100,000 deployed units will tell us which bet was right.