NEO Robot (1X NEO)

How much does the NEO robot cost?

Two tiers, one deposit, and several things the headline numbers don't tell you.

Early Access is $20,000 as a one-time purchase. 1X states this includes ownership, a three-year warranty, premium support and priority delivery.

Standard is $499 a month as a subscription, including the monthly subscription itself, a "Starter Productivity Package" and standard delivery. Note the ordering consequence: 1X states subscription units will be shipped at a later date than Early Access units. Subscribers are behind buyers in a queue that has not started moving.

A $200 deposit is required for either, described as fully refundable. The bounded version, per reporting on 1X's terms, is that it is refundable until 14 days after your "Order Invitation" arrives.

The term almost nobody mentions: the advertised price is not locked at deposit time. 1X's terms allow the price to change before you accept the Order Invitation. You are reserving a place in a queue at an indicative price, not buying a robot at $20,000.

No minimum subscription term is published by 1X. Third-party guides assert a six-month minimum totalling $2,994, but that appears on no 1X page and we could not verify it. Similarly, warranty is contested - 1X's own order page says three years for Early Access, while a third-party guide says one year; subscription-tier warranty terms are not published at all.

Sales tax and delivery are not itemised anywhere on 1X's public pages. Nor is international pricing; deliveries outside the US are stated for 2027 with no figure attached. No price change has been announced during 2026.

What does the $20,000 actually exclude?

Some of this is unknown rather than excluded, and the distinction matters when you're budgeting.

Explicitly not itemised: sales tax and delivery. Neither appears anywhere on 1X's public pages, and no delivery fee, installation fee or setup process has been published. 1X has teased delivery logistics - a video of NEO in a rounded shipping container in April 2026, and a design lead describing the goal as the robot simply showing up and knocking on your door - but no cost or process is public.

Broadband is a genuine prerequisite that carries no published specification. Expert Mode is live VR-rate video streaming from your home to a 1X operator, which needs real upload bandwidth. Third-party guides suggest Wi-Fi 6 and 50 Mbps or more of upload with whole-home coverage. 1X publishes no bandwidth requirement at all, which is itself notable for a product whose headline feature depends on it. NEO without broadband is not a functional product.

Electricity is a rounding error. Third-party estimates put active draw at 200-300 W against an 842 Wh pack with roughly four hours of runtime and a two-hour recharge, working out to somewhere around $5-$20 a month at US rates. Do not let anyone sell you this as a concern.

The subscription's "Starter Productivity Package" implies packaged capability tiers exist above it, though 1X has published no upsell menu and no chore packs. Treat future software tiering as possible but undocumented.

And the most consequential exclusion is information: the binding purchase terms are not public. They arrive with the Order Invitation. You place a $200 deposit before seeing them.

Is a human watching you through the robot?

Sometimes, by design, with your scheduled permission — and 1X is upfront about this rather than hiding it.

The feature is Expert Mode. When NEO cannot complete a task, you schedule a session and a 1X "Expert" remotely pilots the robot, in practice wearing a Meta Quest 3 headset. The operators are 1X employees physically located in the United States. The video from those sessions trains 1X's neural networks - that is the explicit business purpose, not a side effect.

The published controls, all of which are 1X's own claims and none of which have been independently audited: explicit consent per session, so an operator cannot connect without your approval; no-go zones you designate that the software blocks teleoperators from entering; face blurring, on by default, so operators cannot see people's faces; and an indicator light on the robot that changes while a session is live. In autonomous mode, 1X says limited sensor data may go to its servers to fulfil a request but is not stored, and describes sharing data for product improvement as optional.

The gap is retention. 1X has never published a retention period, a deletion mechanism, an audit process or a granular opt-out for teleoperation video. Gizmodo reported in October 2025 that the company retains all video, which sits in unresolved tension with the "optional data sharing" framing. 1X has not reconciled the two publicly.

CEO Bernt Børnich framed the whole arrangement as a social contract, stating that without your data the company cannot make the product better. Whatever you think of that, he is not concealing the trade - he is asserting it as the price of entry.

Our read: the mechanism is honest, the controls are plausible, and the retention terms are the unanswered question. If home video retention matters to you, get the terms in writing before the deposit, not after.

How autonomous is NEO really?

Less than the marketing imagery suggests, and 1X does not actually claim otherwise.

What is autonomous today: walking and locomotion, in-home navigation, answering the door, simple object retrieval, and conversation through an onboard language model. That runs on Redwood AI, published by 1X in June 2025 - a 160 million-parameter transformer operating at roughly 5 Hz on the robot's onboard embedded GPU. 1X's published capability list for Redwood includes retrieving objects, opening doors including bracing a hand against a wall to pull heavy ones, navigating the home, manipulating while moving, picking objects it has not seen before, and bending to pick toys or clothes off the floor.

What is not autonomous: essentially every chore people buy a home robot for. Laundry, folding, the dishwasher, kitchen work, tidying to any standard.

The most useful outside account remains the Wall Street Journal's hands-on published 28 October 2025. The reporter did not see NEO do anything autonomously - every household task in the session was driven by a remote teleoperator. Reporting from the same week recorded the robot nearly toppling over while closing a dishwasher, taking about two minutes to fold a single shirt, struggling to open a refrigerator, and having balance trouble.

1X's own framing is consistent with all of that. Its VP of AI described the model at launch as autonomous with a call for human assistance when NEO cannot do a task, comparing the arrangement to how Waymo is operated and supervised, and warned that the product is early for its time with features still in active development and that there will be mistakes. 1X's product page routes unfamiliar tasks to scheduling an Expert.

1X said at launch it expected NEO to handle most tasks autonomously by 2026 with variable quality. There is no independent evidence either way, because no unit is in a customer's home to test.

Which NEO specs have actually been demonstrated?

This section exists because two of NEO's headline numbers are extraordinary, and neither has ever been shown.

Every figure 1X publishes is a company claim. No production unit has been independently measured by any third party, for the straightforward reason that none is in public hands.

The claims: roughly 1.68 m tall, about 30 kg, an 842 Wh battery giving four hours of runtime with quick-charging at six minutes per hour of runtime, 25 kg carry capacity, 1.4 m/s walking, 8 m/s hand speed, dual 8.85 MP stereo fisheye cameras at 90 Hz, four beamforming microphones, three speakers, 22 dB operating noise, and an NVIDIA Jetson Thor compute module 1X calls NEO Cortex.

Two need flagging hard. The 70 kg - 154 lb - lift figure is on NEO's current spec sheet, not inherited from an older prototype, and it appeared in launch-day coverage. But it is a peak static lift claim from a 30 kg robot, has never been shown on video or measured by anyone, and sits alongside a far more modest 25 kg carry figure. Treat it as an engineering limit, not a usable capability.

The 6.2 m/s running claim is worse. That is about 22 km/h, faster than most humans sprint. It is on the current spec sheet. There is no video, no demonstration and no observation of NEO running at any speed - every public clip shows it walking or standing. Until someone films it, this is a kinematic maximum on paper rather than a capability, and we would not repeat it as one.

The hands changed mid-2026 and 1X's own pages now disagree. A July 2026 post describes 25 degrees of freedom total - 22 actuated in the fingers and palm plus 3 at the wrist - all natively force-controlled and fully backdrivable, with peak torques of 3.5 Nm at the thumb, 2.6 Nm at the finger joints and 17.75 Nm at the wrist, positioning to ±0.2 mm, tactile sensing for normal force, contact location and shear, IP68 sealing, food-safe materials, and wrists tested beyond two million cycles. The main spec page still lists 22 per hand. Both are 1X's pages.

Is NEO safe around children and pets?

1X's engineering answer is genuinely thoughtful. Its own guidance is more cautious than its marketing.

The design argument is mechanical compliance rather than software guarding. NEO weighs about 30 kg - roughly a large dog, and around half the mass of Tesla's Optimus. It wears a soft knit fabric suit in muted colourways with rounded edges. And the actuators are tendon-driven and quasi-direct-drive, meaning they are backdrivable and yield on contact rather than resisting it. That is a materially better safety story than a heavy geared humanoid, and 1X's head of design has framed it as the machine feeling soft because the engineering underneath genuinely is.

But 1X's own FAQ advises staying attentive around children, pets and vulnerable people. In other words, the company does not claim NEO is safe to leave unsupervised with a toddler or a dog. That is honest, and it is also a significant constraint on a product sold as a household helper - if it needs watching, the labour saving shrinks.

On certification there is nothing. No UL listing, no CE or UKCA marking, no ISO 13482 for personal care robots, no consumer-product safety declaration is publicly cited by 1X. For a machine intended to operate unsupervised in private homes at scale, that gap has to close before mass deployment in most markets, and no timeline has been published.

No incident has been reported, which is trivially true since no unit is in a consumer home.

And the liability terms are unknown. 1X's public pages do not surface the insurance or liability allocation in the reservation agreement; the binding terms arrive with the Order Invitation. We are not going to speculate about who is responsible if a remotely piloted robot damages your house, because the document is not public.

Is there a NEO SDK or developer platform?

No. There is no SDK, no developer API, no app store and no third-party skill marketplace as of 30 August 2026.

This one needs a specific correction, because a title has been widely misread. In July 2026 1X published a post called "NEO's Hands - An API to the Physical World." That is a metaphor. The post is a hardware announcement about force-controlled hands. It announces no API, no SDK and no developer access of any kind. Anyone citing that headline as evidence of a developer platform has read the title and not the article.

1X has spoken directionally about capability growing over time, and it founded a World Model Lab in June 2026 as a research organisation, but no dated developer-access commitment exists. The only user-facing extension surface is the mobile app and voice control.

What that means practically: NEO is a closed appliance. You cannot write skills for it, cannot integrate it with your home automation beyond whatever 1X exposes, and cannot extend it if 1X's roadmap does not cover what you want. If a developer platform ever arrives it would be a genuine strategic shift and worth watching for - but do not buy on the expectation of it.

For anyone who wants a humanoid they can program this year, the answer is not NEO and it is not Figure or Tesla either. It is Unitree, whose EDU units ship with a documented SDK, native ROS 2 and a MuJoCo simulator.

What does owning a NEO actually involve?

More unknowns than a $20,000 purchase should carry, and that's the fair summary.

What is published: a three-year warranty and premium support on the $20,000 Early Access tier, over-the-air updates, and 1X's recommendation that maintenance goes through the company. Electricity is trivial - call it $5-$20 a month.

What is not published, and we looked for all of it: repair pricing, service network detail, loaner or replacement policy, turnaround time if the robot breaks, subscription warranty terms, cancellation terms and early-termination penalties, delivery fee, installation fee, and the bandwidth requirement the product depends on. Third-party guides fill some of these gaps with figures we could not verify against any 1X page, including a one-year warranty and a six-month subscription minimum. Where 1X's own page and a third-party guide disagree, trust 1X's page and treat the rest as unconfirmed.

The subscription reportedly includes hardware replacement coverage, which would be the more sensible structure for a first-generation machine - but the terms are not public.

One structural point worth thinking about before either tier. At $499 a month, four years of subscription costs roughly $24,000, more than buying outright. The subscription is not the cheap option; it is the option where somebody else owns the risk on a first-generation product with no service network. Whether that is worth the premium depends entirely on terms that have not been published.

And the ongoing cost nobody prices: your home video. Expert Mode sessions feed 1X's training data by design. That is not a dollar figure, but it is part of what you are paying.

Where is NEO made, and does the July 2026 US robot rule affect it?

Genuinely unresolved, and we are not going to pretend otherwise.

The rule is FCC Public Notice DA 26-786, released 28 July 2026, adding foreign-produced advanced robotic devices to the Covered List. The headline framing of it as a China ban is wrong. The text is country-neutral: "foreign-produced" means the device fails the Buy American domestic-end-product test, which requires US manufacture plus more than 65% domestic component cost. Robots from the UK, EU, Japan, Korea and Canada are caught by the same wording, and US assembly of imported parts does not clear the bar.

The mechanism is narrower than "ban": it blocks new FCC equipment authorisation, and the absence of authorisation is what then blocks import, marketing and sale. It is not a customs seizure order and not a recall. Anything authorised before 28 July 2026 is grandfathered, with import, sale, ownership and resale continuing and a waiver covering firmware and security updates until at least 1 January 2029.

Where that leaves NEO is unclear. 1X is a company with Norwegian origins - founded in 2014 as Halodi Robotics - now headquartered in Palo Alto, with its NEO Factory in Hayward, California, opened April 2026, and legacy manufacturing in Moss, Norway. What 1X has never published is where NEO itself is built or what its domestic content share is. Without that, nobody outside the company can say whether NEO meets the domestic-end-product threshold.

We are stating that as an open question rather than guessing either way. If you are placing a deposit, ask 1X in writing for NEO's FCC ID and the grant date. A grant predating 28 July 2026 settles it in your favour; the absence of one is something you want to know before, not after, you commit.

1X NEO vs Figure 03 vs Tesla Optimus

Three companies, three completely different bets, and the leader changes depending on what you measure.

On ordering, 1X leads and nobody is second. NEO is the only humanoid with an open consumer preorder, a published consumer price, a dedicated consumer factory and more than 10,000 consumer orders. Figure has no price and no preorder. Tesla has no price and told analysts in August 2026 that external sales begin no earlier than H2 2027.

On demonstrated autonomy, Figure leads clearly. Its May 2026 package-sorting run covered roughly 101,000 packages over 81 continuous hours with its CEO stating no teleoperation - a stronger autonomy demonstration than anything 1X has shown. NEO, by contrast, builds remote human operators into the product as a feature.

On real deployment, Figure again. BMW's own June 2026 release places Figure 03 at Plant Spartanburg on logistics sequencing; Tesla has no external site. 1X has an agreement with EQT signed in December 2025 covering up to 10,000 units into EQT's portfolio companies through 2030 - which is an interesting signal in itself: a consumer home robot pivoting toward an industrial channel before a single home delivery.

On delivering a robot into a private home, the score is nil-all-nil. Nobody has done it.

Worth adding to the frame: a counter-thesis is emerging in the shape of cheaper, less ambitious machines that can actually ship - a legless upper-body home robot launched at $7,999 in July 2026 is the clearest example. If NEO's timeline keeps slipping, that category is where the home-robot market may actually begin.

Should you put down the $200 deposit?

It depends almost entirely on how you read the deposit, and we would encourage you to read it as a reservation rather than a purchase.

Put it down if you want to be early, you find the technology genuinely interesting, and you are comfortable that $200 buys a place in a queue at an indicative price that 1X can change before you accept the Order Invitation. It is refundable until fourteen days after that invitation arrives. As an option on being among the first people in the world with a home humanoid, $200 is not an unreasonable price, and there is real engineering behind it - the hands published in July 2026 are legitimately impressive, and Redwood AI is a real onboard model.

Do not put it down expecting a robot to do your chores. Every credible hands-on account describes teleoperated chores, and 1X's own product design routes unfamiliar tasks to a scheduled human operator. If you would not be comfortable with a 1X employee driving a camera-equipped robot through your home, this is not a product you should reserve, and no amount of face blurring changes the underlying arrangement.

Do not put it down expecting delivery this year. There have been zero verified customer deliveries as of 30 August 2026, ten months after preorders opened and four months before the year ends, and 1X's own language has softened from a date to "some this year, some later."

And read the SoftBank situation before you commit. Talks reported on 27 August 2026 involve a majority stake at roughly $6 billion - a down round against the $10 billion 1X sought months earlier, and a control transaction rather than a growth round. That is not a reason not to reserve. It is a reason to keep the refund window in mind.