For a decade, "humanoid robot" meant a lab demo: careful, tethered, rehearsed. In 2026, it means a purchase order. Boston Dynamics' electric Atlas is in production for Hyundai's own factory floors. Agility Robotics' Digit has quietly moved more than 100,000 totes through a real GXO warehouse. Figure AI shipped over 350 of its home-oriented Figure 03 units out of a dedicated factory and signed a paid fleet contract with BMW. And 1X Technologies opened pre-orders on NEO, a robot it is explicitly pitching for your living room, not a loading dock.
That's the headline. The real story is messier and more interesting: almost every company chasing the "robot in your house" dream is actually making its money — and proving its technology — somewhere else entirely. This is a look at the seven companies actually shipping hardware right now, what they're really being used for today versus what they're promising for tomorrow, and why the honest answer to "when do I get a robot butler" is still "not yet, and maybe not the way you think."
The Field
Seven companies, one race, two very different finish lines
Every entrant below is a real, currently operating company with hardware that exists today — no concept renders. But "exists" covers a lot of ground, from robots doing paid shifts in an actual Mercedes-Benz plant to robots that have been pre-sold for home delivery with, as of this writing, no independently verified customer deliveries yet. The cards are ordered roughly by how close each company actually is to a stranger's house, not by hype.
NEO
The most home-focused robot on this list, and the most controversial. NEO is a roughly 5'6" humanoid pitched for chores like laundry, tidying, and answering the door. 1X has been unusually candid that many early tasks are completed via remote human teleoperators rather than onboard autonomy — the company calls its operator console "Starcraft-like," with household occupants blurred from the operator's view and user-defined no-go zones. Critics counter that a live camera-and-microphone feed inside a home is a genuinely new kind of privacy exposure, teleoperator training wheels or not. Pre-orders are open with delivery pledged for 2026; as of mid-2026 no independently confirmed customer delivery had been reported.
Figure 03
Unveiled in late 2025 as Figure's first robot engineered from scratch for the home rather than the factory — wireless inductive charging so it can dock itself, palm-mounted cameras, and fingertip sensors sensitive enough to register a few grams of force, all wrapped around Figure's own "Helix" vision-language-action model built to handle the unpredictability of a real house. The twist: the first real money is coming from manufacturing, not living rooms. By mid-2026 Figure had produced more than 350 units at its BotQ factory and signed what it calls the industry's first paid commercial-scale humanoid contract — an initial fleet for BMW's Spartanburg plant. The home version of the pitch is still mostly a roadmap.
Optimus (V3)
Still the most-hyped, least-externally-verified robot in the field. Tesla wound down Model S and Model X production at its Fremont plant to clear line space for Optimus, with Elon Musk targeting a summer 2026 production ramp and external commercial customers by late 2026 — targets the company itself has already slipped once on the V3 unveil. As of Tesla's own recent earnings commentary, in-factory units remain "primarily for learning and data collection rather than performing productive tasks," and Musk has called near-term production volume "literally impossible to predict" given roughly 10,000 unique parts on an all-new line. No consumer or home version has a real date.
Apollo (Apollo 2)
Backed by Google and Mercedes-Benz, with Google DeepMind's Gemini models doing the reasoning under the hood. Apollo is already doing supervised work inside real Mercedes-Benz manufacturing logistics and GXO Logistics warehouse operations, with a large 2026 funding round pulling in new strategic investors including John Deere and the Qatar Investment Authority. Apptronik opened an expanded "Robot Park" training and deployment facility in Austin and now offers Apollo in both bipedal and wheeled-base configurations depending on the job. No home ambitions have been announced — this is an enterprise play, full stop, at least for now.
Atlas (electric)
The most mechanically advanced humanoid most people have seen on video, now built as a fully electric production robot rather than the hydraulic research platform of years past — 56 degrees of freedom, a 50 kg lift capacity, and dual swappable battery packs it can change autonomously. Parent company Hyundai has committed to deploying more than 25,000 Atlas units across its own Hyundai and Kia plants by 2028, with the first 2026 fleets going to Hyundai's Robotics Metaplant Application Center and to Google DeepMind for further model training. This is a factory robot, not a home robot, and Hyundai has publicly clashed with plant labor unions over the rollout.
Digit
Not flashy, but arguably the most commercially proven humanoid on Earth right now. Digit's distinctive reverse-knee legs are purpose-built for warehouse tote-moving, and the numbers back it up: over 100,000 totes moved at a single GXO facility, tens of thousands of paid operating hours, and robots-as-a-service contracts running with GXO, Toyota, Mercado Libre, and Schaeffler. Agility is preparing a public listing and just opened a new Physical-AI development hub in Fremont. There is no consumer or home version of Digit, and the company has not signaled one is coming — its bet is that warehouses, not living rooms, are where humanoids pay for themselves first.
G1 & R1
The volume play. Unitree claims the top spot in global humanoid shipments, built on a research- and developer-oriented G1 platform sold in a dozen configurations, plus the smaller, child-height R1 — a deliberately less imposing humanoid the company markets as "less intimidating" for home settings and easier to move around a house. Unitree's public demos, including a televised fleet of G1s performing a fully autonomous kung fu routine in early 2026, lean hard into choreography over open-ended chores. The company is targeting a roughly fourfold jump in annual production for 2026. Genuine general-purpose home autonomy has not been demonstrated at anywhere near the level of its stage routines.
The Real Divide
Almost every "home robot" is actually a factory robot in disguise
Line the seven up and a pattern jumps out immediately: the companies with the most believable near-term traction — Agility, Apptronik, Boston Dynamics — aren't chasing homes at all. They're chasing warehouses and assembly lines, where the work is repetitive, the environment is controlled, and a robot's mistake costs a dented tote rather than a broken lamp or a startled toddler. Digit's 100,000-tote milestone and Apollo's Mercedes-Benz pilot are real, audited, paid deployments. That's precisely why they're industrial: those are the environments general-purpose robots can actually handle today.
The companies explicitly promising your living room — 1X and, in its marketing if not yet its revenue, Figure — are running a harder experiment. A house has irregular lighting, pets, stairs, breakable objects, and zero tolerance for a 130-pound machine falling over near a child. That gap is exactly why 1X leans so heavily on human teleoperators rather than claiming full autonomy, and why Figure's actual paid contract to date is a BMW factory fleet, not a homeowner subscription. The honest read: home-capable hardware is arguably closer than home-capable software and home-acceptable trust.
The Uncomfortable Part
The teleoperation question nobody's fully answered
1X deserves credit for being unusually upfront about how NEO actually completes its more complex tasks today: a trained human operator, working remotely, drives the robot through the motion while it also gathers data to eventually reduce that dependence. The company's stated safeguards are specific — occupants blurred out of the operator's video feed, operators unaware which household they're connected to, one supervisor overseeing roughly eight teleoperators, and homeowner-defined zones the robot won't enter even under remote control.
Why it matters
Whatever the safeguards, a live camera-and-microphone feed originating inside a private home, routed to a person the homeowner has never met, is a materially different privacy proposition than a robot vacuum bumping into furniture. 1X has argued supervised teleoperation is more auditable than an unmonitored in-home cleaner, since every session is logged. Reasonable people, including robotics researchers covering the launch, have disagreed on whether that argument holds up once you're an early customer rather than a case study.
None of this means NEO is a scam, or that teleoperation-assisted robots are dishonest — it's a legitimate, disclosed bridge strategy while onboard autonomy catches up, and it's arguably more transparent than competitors who haven't detailed how their own demos are actually being run behind the scenes. But anyone evaluating a home humanoid in 2026 should ask, specifically, how much of what they're seeing in a promotional video is the robot deciding, versus a person deciding for it.
Reality Check
What "2026" actually means for this category
Safety and regulatory frameworks haven't caught up to the hardware. As of 2026, there is no single, finished safety standard governing humanoid robots operating around untrained people in a home; the relevant international draft standard is still working its way through committee, with formal publication expected in 2027 or later, and it's not expected to include enforcement mechanisms even then. In the U.S., oversight remains fragmented across existing consumer-product and workplace-safety rules never written with a bipedal machine in mind. The EU has been faster to move on certification requirements, but nothing resembling a mature, universally recognized "home robot safety" standard exists yet anywhere.
Take that alongside the actual deployment picture — paid, supervised, industrial work is real and growing; unsupervised general-purpose home autonomy is not proven at scale by anyone on this list — and a fair summary for 2026 looks like this: this is the year humanoid robots proved they could hold a job. It is not yet the year they proved they could be left alone in your house.
Reporting note: This report was independently researched and written by DopamineKart's editorial desk. No company named above reviewed, sponsored, or paid for this coverage. Figures on deployments, funding, and timelines reflect public company statements and reporting available as of August 2026 and are subject to change as these programs evolve — several of the products described (notably NEO and Optimus V3) are pre-release or early-access hardware.
Spot something that's changed or gotten stale? Corrections go to corrections@dopaminekart.com.